Features & Improvements

• Organize complex automations effortlessly with new Workflow tools, featuring node grouping, list support, and visible Compute Unit (CU) costs.

• Drive video and image generation with the newly upgraded Compositor node: a full video editor inside your node editor.

• Prevent generation errors early via proactive warnings when input assets exceed a model's maximum size constraints.

• Navigate enterprise integrations easily on the updated Bring Your Own Key page, featuring clearer naming and reorganized info panels.

Fixes

• OBJ files generated within workflows are now properly bundled as ZIP archives for consistent 3D asset downloads.

• Experience smoother platform navigation, enhanced team usage heatmap, refined boolean toggles for custom interfaces, and streamlined app deletion.

Features & Improvements

• Workflow nodes: Build complex visual automations using the newly introduced Compositor nodes.

• Enterprise Key Management: Manage external integrations securely with Bring Your Own Key (BYOK) support.

• Team model access: Control exactly which models your team can use by toggling between allowlist and blocklist modes.

• Platform polish: Discover the Tripo P1 3D model faster via a new homepage highlight. Additional UI updates include refined branding, a theme-aware favicon, and clearer team invitation tracking.

Fixes

• Workflows properly ignore deleted assets in file arrays, preventing execution errors.

• Duplicated nodes no longer retain memory of previous connections or data from their original source.

• Enhanced overall platform stability and resolved minor UI alignment issues.

Tripo P1: Production-grade native 3D diffusion model that generates engine-ready assets directly in 3D space with clean topology and stable geometry. Supports PBR materials, texture quality control, face limits, and automatic real-world scaling.

Tripo P1 MultiView: Multiview variant of the Tripo P1 model that accepts front, left, back, and right view images to produce a single 3D mesh. Ideal for higher-fidelity asset reconstruction from multiple reference angles.

HeyGen Suite

HeyGen Video Translate Precision: High-quality video translation at 10 credits per output second. Supports over 100 languages and dialects, with options for audio-only translation and dynamic duration adjustment for natural pacing across languages.

HeyGen Video Translate Speed: Faster video translation at 5 credits per output second. Shares the same broad language support as Precision mode, optimized for speed and high-volume workflows.

HeyGen Video Agent: AI-powered explainer video generation from natural language prompts. Supports portrait and landscape orientations, a large library of avatar characters, and durations from 5 to 120 seconds.

HeyGen Avatar 4: Creates realistic talking avatar videos from a photo and text prompt. Supports built-in TTS voices, resolution presets up to 1080p, stable or expressive talking styles, and optional caption overlays.

Voice and Audio

Lux TTS: High-quality voice cloning TTS at 48kHz from text and a reference audio clip. Supports guidance scale control for adherence to the reference voice and adjustable inference steps.

Tada 1B: Lightweight voice cloning text-to-speech model with multilingual support. Offers controls for speech speed, sampling temperature, and acoustic guidance.

Tada 3B: Higher-capacity voice cloning text-to-speech model with multilingual alignment and expressive controls for more detailed and nuanced voice output.

Specialized Tools

VecGlypher Image to SVG: Generates SVG glyphs from a text prompt while matching the style of reference glyph images. Supports 1 to 8 reference images, customizable fill and stroke colors, and output sizes up to 4096px.

Physic Edit: Physics-aware image editing that applies realistic refraction, material changes, and deformations to an input image based on a text prompt.

Pixa Background Removal: Fast, ultra high-quality background removal from images. Supports RGBA and alpha output formats. Well suited for e-commerce and image editing workflows.

Workflow and Node Editor Upgrades

Quick-Connect Menu: Dragging a connection to an empty area now opens a filtered context menu, letting you create and connect a new node in a single motion.

Multi-Select: Use Shift + Drag to select multiple nodes and edges for bulk moving or editing.

Input Management: Improved the reordering UI for workflow inputs with numbered badges and clearer tooltips.

Batch Export: Added a "Download All" button to model nodes with customizable format options.

Improved Model Training (Musubi)

Better Defaults: Musubi training now features optimized defaults and refined image limits for more consistent results out of the box.

Enhanced Progress: The progress bar now tracks days for elapsed and remaining time, providing better clarity for long-running jobs.

Limit Awareness: The image counter now displays your plan-specific limits and includes a direct prompt to upgrade if you need more capacity.

Team and Enterprise Management

Team Dashboard: Fixed an issue where project owners were restricted from the team dashboard. They now have full access.

Admin Controls: The Terms and Conditions modal is now shown only to team admins to reduce friction for general users.

Team Scaling: Added a page size selector to the team members panel for easier navigation in large organizations.

Enterprise Filtering: You can now filter models by Enterprise Status to quickly find approved assets.

Bug Fixes and Refinements

Live Canvas: Fixed filtering to ensure only compatible models are visible.

Tool Redirection: Resolved an issue where "Use Model" on Vectorize and Background Remove sent users to the wrong generation page instead of the specific tool.

Clipboard Reliability: Improved error handling for clipboard copies when browser permissions are blocked.

App Logic: Fixed issues with renaming and creating apps on newly generated workflows.

This release expands Scenario's 3D, video, voice, and image editing capabilities while improving workflow usability and platform stability. You can learn more about these features in our Knowledge Base.

LTX Series

LTX 2.3 Pro: High-fidelity video generation model designed for cinematic outputs and complex motion synchronization. Supports first and end frame inputs, resolutions up to 4K, adjustable clip durations, and multiple camera motion options.

LTX 2.3 Fast: Speed-optimized video generation model designed for rapid iteration and high-volume production. Generates videos quickly while maintaining synchronized audio.

LTX 2.3 Pro Extend Video: Extends existing footage while preserving style and structure. Generates continuation frames from an input video to build longer sequences.

LTX 2.3 Pro Retake: Allows precise editing of existing videos by regenerating selected segments. Supports replacing audio, video, or both within a defined portion of a clip.

Read more about the LTX 2.3 series in our knowledge base.

Kling V3 Series

Kling V3 Omni Video: Video model supporting text-to-video, image-to-video, and video-to-video workflows. Enables generation from prompts, transitions between start and end images, or restyling using reference videos.

For more info about the model, check our our knowledge base article.

Kling V3 Std Motion Control: Motion transfer model designed to animate static character images using movements from a reference video.

Kling V3 Pro Motion Control: High-fidelity motion transfer model optimized for complex gestures and detailed character animation.

You can read more about the Kling V3 Motion Control series here.

3D Tools

Tencent Texture Edit: AI-powered texture editing for 3D models. Applies textures from text prompts or reference images to FBX models and supports PBR materials when using prompts.

Trellis 2 Retexture: Retextures existing 3D models using a reference image to generate new texture maps. Supports FBX, OBJ, and GLB meshes with high-resolution texture outputs. You can read more about the model here.

Workflow & Developer Experience

Sticky Notes in Workflows
You can now add Sticky Notes directly to workflows to document logic or leave instructions for collaborators.

Video Frame Selection
Video model nodes now allow explicit selection of first and last frames for precise motion interpolation.

API Parameter Visibility
Model pages now display API parameters to make it easier to integrate Scenario models into external applications.

Relaunch Failed Generations
Failed generations can now be relaunched directly.

Retraining Flow Update
When retraining a model, users are now automatically redirected to model selection.

Fixes & Quality of Life Improvements

Upload Processing Fix
Resolved an issue where 3D and video uploads could occasionally get stuck during processing.

Preview Model Status
Models that are experimental or in early access now display a Preview tag.

Creative Tool Stability
Fixed issues affecting Prompt Spark and Explore when a LoRA was selected.

Resolution Parameter Fix
Selecting a resolution preset no longer clears custom width values.

This release expands Scenario’s video and 3D capabilities while improving workflow usability and platform stability. You can learn more about these features in our Knowledge Base.

More updates and new releases are coming soon.

Tripo 3.1: Generates high-quality 3D models from image inputs with improved texture fidelity and PBR support. Designed for cleaner topology and production-ready assets.

Tripo Rigging 2.5: Automatically prepares 3D meshes for animation by generating skeleton structures, skin weights, and rig setups. Supports multiple rig types including bipeds, quadrupeds, hexapods, serpentine creatures, and aquatic forms.

Tencent UV Unwrapping: Automatically generates clean UV maps for existing 3D models. Supports FBX, OBJ, and GLB formats and handles meshes with up to 30,000 faces, making it easier to prepare assets for texturing and material workflows.

Video Generation

P-Video: High-speed video generation model designed for rapid prototyping and social content. Generates clips up to 1080p with durations between 1-10 seconds, supporting 48 FPS motion, seven aspect ratios, and a Draft Mode for faster previews.

Model Training

You can now train custom LoRAs using several next-generation base models:

• Flux 2 - Open-source image generation models from Black Forest Labs with strong prompt adherence.

• Flux 2 Edit - Image editing model trained on before-and-after pairs for precise changes such as style shifts or object swaps.

• Qwen Image - Text-to-image model from Alibaba capable of generating diverse styles with strong prompt alignment.

• Qwen Image Edit - Image editing model designed to modify scenes while preserving characters and composition.

• Z Turbo - Fast image generation model optimized for strong training fidelity and high-quality outputs.

These models allow you to train custom LoRAs tailored for image generation or editing workflows, depending on the base model you select.

Workflow & Infrastructure Improvements

Concurrency Management
Job and workflow limits were introduced to improve platform stability and ensure fair resource distribution across users.

Node Editor Enhancements
Model nodes can now output multiple results simultaneously, enabling more advanced branching and complex workflow logic.

Smarter Error Messages
Failed jobs now include descriptive hints to help identify the cause and resolve issues faster.

Enterprise Branding Updates
Model License labels have been replaced with a clearer Enterprise Status indicator.

Fixes & Quality of Life Improvements

Improved Error Management
Upgraded internal error handling for clearer reporting.

Prompt Persistence Fix
Resolved an issue where switching models would erase text in Edit with Prompt.

Node Logic Fixes

• Fixed an issue where If/Else nodes incorrectly flagged image lists as empty

• Resolved stale image references in LLM nodes

• Fixed the jumping cursor issue in the Prompt Builder

Enterprise Billing
Corrected a display bug where the Current Plan price appeared incorrectly for Enterprise users.

Training Flow Optimization
The plan restriction modal now appears at the end of the training setup process, allowing users to complete configuration before plan compatibility is checked.

SVG Workflow Downloads
Fixed an issue that prevented SVG files from downloading from workflows.

This release expands Scenario’s capabilities across 3D generation, video creation, and custom model training, while improving workflow flexibility and platform stability. You can learn more about these features in our Knowledge Base.

More models and workflow enhancements are coming soon.

3D Generation

Tripo 3.0: Generates high-fidelity 3D meshes from text or image prompts. Includes PBR map generation, quad mesh optimization, adjustable face limits for poly-count control, and Standard or HD quality modes.

Tripo 3.0 Multiview: Builds more structurally accurate models using four reference angles (front, left, back, right). This removes visual blind spots and keeps proportions consistent from every perspective.

Tripo 3.0 Texturing: Creates high-resolution PBR materials for existing models using a text prompt, image prompt, or style reference.

You can read more about the Tripo 3.0 suite on our knowledge base.

Image Generation

Recraft v4: Raster image generation model designed for clean composition and natural language prompting. Optimized for web visuals and rapid ideation.

Recraft v4 Pro: Higher-detail version with improved lighting and texture quality. Suitable for professional visuals and high-resolution publishing.

Recraft v4 SVG / Pro SVG: Generates true vector graphics with editable paths and layers instead of rasterized images. Ideal for UI elements and simple logos.

For more info about the Recraft v4 suite, check out this article.

Video Generation

LTX-2 19B: Generates video from text or images with synchronized audio created as part of the same process. Provides high visual fidelity and advanced controls such as guidance and diffusion steps.

LTX-2 19B Fast: Speed-optimized version for rapid iteration and previews while maintaining motion consistency. Useful for quickly testing ideas before final renders.

LTX-2 19B Extend Video: Continues an existing video by generating new frames that preserve motion and scene continuity. Designed for extending clips naturally rather than restarting a generation.

You can read more about the LTX-2 19B Models here.

Runway Gen-4.5: High motion quality and strong prompt adherence for cinematic video generation.

Workflow & Canvas Enhancements

We added tools that make workflows easier to manage and more precise to use.

Export & Import Workflows: You can now export workflows and import them into another project or account.

Mini Canvas Text: You can place text notes directly on the canvas, useful for instructions or Edit-with-Prompt markups.

Workflow Locking: Lock a workflow to prevent accidental changes to node structures.

Video Node First/Last Frame: Video nodes now support first and last frame inputs, giving more control over transitions and consistency.

Custom Vectorization: The Vectorize tool now runs using custom models.

Fixes & Quality of Life

Multi-Image Import: Multi-image nodes now correctly batch imported assets.

Analytics Datepicker: Updated date selector interface for smoother use in Analytics.

This release continues expanding what you can create directly inside Scenario while making workflows more controllable and easier to manage day to day. More improvements and models are already on the way.

Image Generation & Editing

xAI Grok Imagine Image: Generates and edits images from text or an existing image, with strong prompt accuracy and support for multiple aspect ratios and batch outputs.

xAI Grok Imagine Image Pro: A higher-fidelity version designed for detailed and production-quality results, with higher resolution outputs and improved editing performance.

Video Generation

xAI Grok Imagine Video: Generates video from text, images, or an existing video. Supports up to 15-second clips, configurable aspect ratios, and synchronized audio including dialogue and sound effects.

Features & Improvements

Workflow Modes: Improved behavior of Read, Edit, and App modes for clearer workflow navigation.

LLM Node Integration: Outputs from the LLM node can now be used directly inside the prompt builder.

Analytics Export: You can export the “Top Users” section in Analytics as a CSV file.

Tool Nodes: Fixed missing output handles on tool nodes so connections behave correctly.

Bug Fixes

We focused heavily on workflow reliability and node behavior.

• Fixed reference issues affecting duplicated nodes.

• Fixed duplicated nodes resizing together with the original node.

• Fixed multiple image inputs where only the first image was used by a model node.

• Fixed text inputs not functioning correctly inside the LLM node.

• Fixed duplicated workflows sometimes missing from the list or missing thumbnails.

• Fixed the default text-to-model connection reappearing after deletion.

• Fixed double workflow creation when saving quickly with Cmd/Ctrl + S.

As always, we will continue bringing new models and ongoing improvements across the platform. You can also read our Knowledge Base to learn more about these features and how to use them in your workflows.

You can now build a full end-to-end process all at once using nodes, run it whenever you need, and even share it with others as a simple App.

Workflows

Scenario V2 introduces Workflows, a visual node-based editor that lets you connect models, tools, and assets into a single creative pipeline. Instead of generating, exporting, opening another tool, and repeating steps, you can now design the full process once and run everything together.

Workflows let you build visually. Each node represents a step such as a prompt, a generation, a transformation, or an output. Connect them and Scenario executes the entire chain for you.

This turns multi-step creative work into a repeatable system rather than a manual sequence of actions.

Building and Editing

We focused on making the editor comfortable to use and safe to experiment in.

Undo and Redo
You can now revert or re-apply actions while building a workflow.

Copy & Paste Across Workflows
Nodes can be copied and pasted between workflows so working setups can be reused instead of rebuilt.

Multi-Asset Nodes
Nodes now support multiple input assets, allowing you to process several images or files inside a single pipeline.

Keyboard Shortcuts
You can now duplicate a node with Cmd/Ctrl + D and copy or paste nodes using Cmd/Ctrl + C and Cmd/Ctrl + V to quickly reuse parts of a workflow.

Smart Workflow Saving
Workflows are only created after you actually modify them, preventing a long list of empty “new workflow” entries.

Workflows and Apps

Workflows can now directly power Apps, and you can move between the two freely.

Workflows → Apps
Any workflow can be turned into an App. Complex pipelines become simple one-click tools that can be shared with other users.

Apps → Workflows
You can open an App and access its workflow to understand how it works or customize it.

Navigation and Controls

We also improved how workflows are accessed and organized.

• Read, Edit, and App viewing modes for workflows

• “After” thumbnails now appear on workflows and show on hover

• Persistent project and platform selection in workflow and app lists

• Sort workflows and apps by last modification date

New Capabilities

LLM Node
A language model node allows text generation and language-based logic directly inside workflows.

Analytics Dashboard
A new analytics dashboard with an activity trail helps you track actions and usage.

Training Comparison
The prompt is now displayed in the epoch comparison modal.

Project Description
Projects now support a description field.

• Usage API: enhanced usage API to better track model usage and global entities

• Workflows API: general improvements and refinements

• Search API: performance improvements for faster and more powerful search

Bug fixes 🐛:

• Assets: audio assets can now be downloaded in bulk

New Frontier Models

(i) Sparc 3D 2.0

Sparc 3D 2.0 is a specialized model from Hitem3D designed for high-fidelity 3D asset generation, succeeding the previous Sparc 3D 1.5 variant. As the latest evolution, this new model addresses key challenges in generative 3D workflows such as:

• Intrinsic Albedo: Custom de-lighting strips baked-in shadows, making assets easier to relight in any engine.

• Unified PBR: The new "1536 pro" model follows geometry tightly; less “texture stretching” on complex limbs.

• Realistic Anatomy: Expect high-fidelity proportions or hair patterns that grow naturally from the mesh structure.

Documentation here.

(ii) Qwen-Image-2512

Qwen Image 2512 is Alibaba’s latest open-source text-to-image model, released in Dec 2025. It delivers high-resolution, photorealistic images with strong prompt adherence, precise spatial composition, and clean text rendering. Optimized for image quality over speed (~6s), it reduces common AI artifacts, making it ideal for character art, product visuals, and lots of diverse marketing content.

Check out the “pinned example” section for more details.

(iii) Flux 2 Turbo Edit

The Turbo variant of Flux 2 by Black Forest Labs is built for speed and efficiency. At just 2 CU, it enables near-instant image edits and rapid iteration. Ideal for simpler adjustments, it lets you test dozens of variations in the time of a single high-quality run - at 10% of the cost of Nanobanana Pro. More documentation here.

New Image Edition Tools

We’ve expanded Scenario’s image editing capabilities with a new Tools tab in the Models page. Navigate to Models → Tools 👉 https://app.scenario.com/models/list?tab=tool

You now have access to 20+ new image editing tools in the Tools tab, allowing you to tweak, clean up, or restyle an image after it’s been generated -faster and more cost-effective than running new prompts.

These tools include Color & Lighting controls such as Color Correction, 3D Color LUT, Tint, Desaturate, Solarize, as well as Dodge & Burn, Glow & Bloom, Grain, and Vignette. You also get a range of Style & Effects tools to further transform your images, including Posterize, Crystallize, Cubism, Oilify, and Parabolize. Additional effects like Chromatic Aberration, Blur, and Dissolve help you fine-tune mood and visual impact with just a few clicks.

“Quality of Life” Improvements

We've implemented several features to make the Scenario experience more efficient and intuitive.

Asset Management

1. Search by ID: You can now search for assets, models, and workflows directly by their unique ID for faster and more precise retrieval.

2. FBX Export: We've added support for the FBX format when downloading 3D models.

3. Derived Image Count: The number of derived images is now displayed in the aside section title.

Workflow Enhancements

1. Video Audio Preview: Audio now plays automatically when you hover over videos, making it easier to preview your video assets.

2. “Clear All” Filters: A new single-action button lets you clear all active filters at once, saving time and clicks.

3. Zoom on Reference Images: You can zoom in and out after opening a reference image for sketching, using controls at the bottom of the modal.

4. Empty State UI: We’ve added helpful empty-state UI in the “Generate 3D” and “Generate Audio” sections to guide users when no content is present.

5. Direct Model Access via Task Manager: You can now open a model directly from the task manager (“bell icon”) once an upload or training is complete.

Bug Fixes

Over the holidays, we've squashed a number of bugs to improve stability and provide a more reliable experience.

UI & UX Fixes

• Added an empty state message when a user is not found within an Organization.

• Jobs now display progress indicators when available

• Reference image sections now automatically open when you paste content

• Adjusted the column sizing for the project selection name field

• Updated terminology in the asset transfer modal

• Addressed and fixed various filtering issues on the model page

Model & Asset Fixes

• Model header tags now correctly reflect the model type.

• Removed the "Expand" option for textures where it was not applicable.

• Fixed an issue with missing thumbnails for 3D nodes.

• Model statistics are no longer displayed when the model is not owned by the user.

• Private LoRAs are no longer shown in Flux Kontext Edit when switching projects

• 3d: added support for Trellis.2

• Detection: improved scalability

• Training: prepare for new training, such as Qwen edit, Qwen edit plus and Z-image base models

Bug fixes 🐛:

• 3D generation: fixed an issue in Hitem3d processing that caused a generation loss.

• Image generation: fixed an issue in Gemini leading to no proposed generation. The endpoint now automatically retries up to 3 times when the problem is not related to Google’s filtering.

New AI Models

Seedance 1.5 Pro is ByteDance’s latest video generation model, capable of producing 1080p clips with first- and last-frame control, consistent motion, and natively integrated audio from text or image inputs. Key capabilities:

• Very cost-efficient - 12 to 48 credits only on Scenario (Dec 25). A true bargain for the output quality.

• Native Audio-Visual Generation - Video frames and audio waveforms are generated simultaneously, ensuring precise synchronization (e.g., footsteps landing exactly on the correct frame).

• Multilingual Lip Sync - Accurate dialogue and facial animation in English, Mandarin, Spanish with support for non-verbal cues (sighs, laughter…).

• Directorial Control - Understands cinematic terminology (e.g., dolly zooms, long-take tracking shots) to maintain character and scene consistency across 4–12 second clips

Be sure to check the “Pinned Examples” on the model page, and the “prompt spark“ for faster prompting!

Kling 2.6 Motion Control

Take command of cinematic movement with Kling 2.6 Motion Control. This model from Kuaishou provides granular control over motion dynamics, making it easier to achieve professional-grade video results.

MMAudio 2 – Add Audio to Silent Video

Produce original, high-quality audio directly within Scenario. Whether you need sound effects, atmosphere, or musical elements, MMAudio 2 provides control over your auditory output:

• MMAudio 2: https://app.scenario.com/models/model_mm-audio-2

• Text to Audio: https://app.scenario.com/models/model_mm-audio-2-t2a

Pika 2.2 Suite

We are integrating the last generation of Pika models, with consistency in video generation and scene composition.

• Pika Text-to-Video: https://app.scenario.com/models/model_pika-2-2-t2v

• Pika Image-to-Video: https://app.scenario.com/models/model_pika-2-2-i2v

• Pika Scenes: https://app.scenario.com/models/model_pika-2-2-scenes

• Pika Frames: https://app.scenario.com/models/model_pika-2-2-frames

New Editing Models

Check out the new “Edit with Prompts” interface on https://app.scenario.com/edit-with-prompts

To the top-left, you will find a “More Models” option to load a series of new editing models like Flux 2 (Pro, Max, Flex) or LongCat Image, a new specialized model for seamless image manipulation and editing, expanding the toolkit for refining images.

We also added Qwen Image Layered, which generate images with separate foreground, background, and shadow layers for easy editing and compositing. More documentation here.

Features & Improvements 🚀

Improved 3D Viewer

Our 3D viewer has been optimized for better performance and visual fidelity, providing a smoother experience when inspecting meshes and textures.

More Enterprise Model Filters (Governance)

For our Enterprise users, we’ve refined the filters on the Models Settings page. You can now filter models based on compliance data, ensuring your team stays within governance requirements at a glance.

Project Selection Modal

Managing multiple projects is now simpler: For Organizations with more than 5 projects, we’ve introduced a dedicated Project Selection Modal to make it easier to navigate between Teamspaces (Projects)

Tag Search Weighting

Searching by tags is now smarter. We’ve improved the weighting of tag-based results to ensure the most relevant models and assets appear at the top of your search.

Edit with Prompts - Expanded Model Selection

We’ve added a "More" option to the “Edit with Prompts” interface. This makes it easier to browse and select from our larger range of editing models. Expect more editing models to be added in the near future!

Bug Fixes 🐛

• CU Calculations: Fixed issues where Compute Unit (CU) calculations were incorrect for certain file array inputs with minimum length requirements.

• Flux Size Enforcement: Ensured that generations using Flux models strictly respect size requirements by enforcing dimensions in multiples of 16.

• Filter Layout: Added horizontal scrolling to the generation filters to prevent layout overflow on smaller screens.

• Background Removal: Updated the "Remove Background" behavior to force shadow removal when used within the reference image section, resulting in cleaner cutouts.

• 3d rendering: enhanced studio setup for better mesh previews

• Images and Videos: added support for post-processing, supporting 16+ distinct effects:

  • 3D LUTs Integration: Added native support for Look-Up Tables (LUTs).

  • Film Simulation: Added realistic grain engine with profiles, halation, and bloom.

  • Color Grading: Full color correction suite (exposure, temp, tint, etc.).

  • Artistic Filters: Vignette, Posterize, Solarize, Crystallize, Oil Paint (Kuwahara), and more.

  • Video Support: Frame-by-frame processing with audio preservation.

Before	After (LUT Matrix v2)

• Various bug fixes and improvements

• Assets: fixed an issue where the image orientation wasn’t taken into account properly, leading to assets with an inverted width and height metadata.

• Models: Improved GET

• Video Generation: fixed an issue with wan 2.6, leading to assets not taken into account

Image & Upscaling

OpenAI - GPT Image 1.5

This model follows complex instructions more accurately than previous versions, making it easier to generate or edit images with precision. It can render clear, readable text for logos and menus, and its "Semantic Locking" feature lets you change specific details (like a character's clothing) without altering the rest of the image. It is also 4x faster, allowing for quicker changes and variations. For a practical overview, see The Essentials guide, or try the model here on Scenario: GPT 1.5 Model Page.

Magnific Upscalers (Enterprise Plans Only)

A well-known high-fidelity upscaler - adds high-resolution detail and texture to any image by "reimagining" missing pixels rather than just stretching them.

Black Forest Labs

Flux 2 Max: This new version of Flux 2 introduces "Multi-Reference" support, allowing you to use up to 10 images at once to keep characters or styles consistent across new generations. It handles complex materials like glass and fabric with higher realism and supports exact brand colors through hex code matching.

Video Models

Wan 2.6 (T2V & I2V)

Cinematic video with high motion stability.

• Wan 2.6 Image-to-Video: https://app.scenario.com/models/model_wan-2-6-i2v

• Wan 2.6 Text-to-Video: https://app.scenario.com/models/model_wan-2-6-t2v

Veo 3.1 "Extend Video"

Lengthen existing clips while maintaining temporal and visual consistency.

https://app.scenario.com/models/model_veo3-1-extend-video. Limitations to be included in KC: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/video/extend-a-veo-video#limitations

New Generative 3D Models

• Meta SAM3D Suite: Precise object isolation and spatial alignment for 3D workflows.

  • SAM3D Objects: https://app.scenario.com/models/model_meta-sam-3d-objects

  • SAM3D Align: https://app.scenario.com/models/model_meta-sam-3d-align

  • SAM3D Body: https://app.scenario.com/models/model_meta-sam-3d-body

Microsoft Trellis 2: High-quality geometry generation for rapid 3D prototyping.

Tools & Editing

• Meta SAM3 (Images & Video): Surgical precision for object and character isolation.

  • SAM3 for Images: https://app.scenario.com/models/model_meta-sam-3-image

  • SAM3 for Videos: https://app.scenario.com/models/model_meta-sam-3-video

• Photoroom Tool Suite: Native professional editing tools.

  • Remove Background: Industrial-strength isolation (includes Bria models).

  • Background Replacer: Instant environment swaps with consistent lighting.

  • Relighting: Modify the mood and light sources of generated images.

  • Text Remover: Clean up artifacts or unwanted text.

  • Uncrop & Expand: Extend the canvas in any direction.

Governance & Enterprise Readiness

We’re making Scenario the most secure and manageable platform for professional teams.

• Model Management & Compliance: Visualize compliance data directly within model details. Track provenance and governance metadata at a glance.

• Editable Metadata: Owners can now edit short descriptions for their models to improve searchability and organization.

• Public Apps & Feed: The Apps page is now publicly accessible! We’ve also improved discovery with a revamped search across the public feed for assets and models.

Workflow Improvements & Shortcuts

We’ve shaved seconds off every action to keep you in the flow.

• Batch Action Shortcuts: New lightning-fast shortcuts to Tag, Delete, Add to Collection, or Download assets instantly.

• Session Filters: Find the perfect generation faster with new search filters available directly within your sessions.

• Custom BG Flows: More models are now compatible with our custom "Remove Background" workflow.

• Usage Calendar: Updated to start on Monday for better planning.

Bug Fixes & Stability

• Thumbnail Reliability: Fixed missing thumbnails in the asset zoom view and model management dashboard.

• Inference Visibility: Textures and skyboxes now correctly show "running" states during generation.

• Cost Clarity: Fixed the CU cost display for Minimax Music models.

• Workspace Privacy: Recent Creations no longer show hidden assets or assets from other users in the workspace.

• Search Accuracy: Fixed issues where upscale capabilities didn't show the correct model count or return specific models by name.

This is the most powerful version of Scenario ever built. Log in now to explore the new models and tools!

👉 Explore the New Models: https://app.scenario.com/models

• Jobs: added Cost Units details to the Jobs’ API. An api user can now access the cost “plain or discounted” of a given job

• Models: added support for short descriptions, helping users and LLM to discover them more easily

• Workflows: added support for UI config in inputs

• Workflows: added support for prompt spark in the runtime

Bug fixes 🐛:

• Assets: fixed an issue when a job was creating a large number of assets at a time (~ 20 or more assets)

• Assets: fixed a missing compression on png files

• Models: more accurate errors when the model’s provider is Fal

• Projects: improved GET requests

• 3d: fixed an issue with certain 3d model providers leading to un readable properties

• Websockets: added information to ws so the reader can distinguish irrelevant assets (intermediary assets, for example) from actual assets with users’ interest more easily

• 3d Generation: added support for SAM 3D

• Image Generation: optimized support for `gpt-image-1` + added support for gpt-image-1.5

• Video Generation: added support for veo3 extension

• Training: improved scalability

Video Models

Kling O1: This model expands Scenario’s video capabilities with multiple creation and editing modes focused on cinematic consistency and control.

• Reference Video: Generate new shots guided by an existing video while preserving motion, camera style, and scene continuity.

• Video Editing: Edit an existing video using natural-language instructions while keeping the original motion structure intact.

• Reference Images: Turn reference images into consistent video scenes with stable character identity, objects, and environments across frames.

Kling AI Avatar v2 Pro: Create high-quality talking avatar videos featuring realistic humans, animals, cartoons, or stylized characters, designed for expressive motion and polished results.

Sync Lipsync React-1: Refine facial expressions and generate realistic lip-sync while preserving visual details in video outputs.

3D Models

Meshy Rigging
Automatically rig 3D models to prepare them for animation, reducing manual setup and speeding up character workflows.

Audio Models

Minimax Music 2
Generate original music tracks directly in Scenario, adding sound and atmosphere to your creative projects.

Image Editing Models

P-Image Edit: A new image-editing model focused on precise visual adjustments and controlled transformations.

Qwen LoRA - Apply Vinyl Wrap: Apply realistic vinyl wrap effects to objects, making it easy to visualize surface materials.

Features & Improvements

Animated GLB Management: Improved handling and preview of animated GLB assets, making it easier to work with animated 3D content.

Multi-Image Drag & Drop Selection: You can now reorder reference images by dragging and dropping them to control how inputs are interpreted.

Nested Inputs for Kling O1: Added structured, nested inputs to provide more precise control over Kling O1 video generation parameters.

Renamed Tags Category: Updated naming for model tags to improve clarity and organization across the platform.

Video to Image Generation: You can now generate still images directly from videos, making it easier to extract frames or reuse video content.

Apps Asset List Visibility: You can now hide App-generated results from your asset list to keep your workspace cleaner.

Bug Fixes

• Fixed glitches in the Apps creation menu and improved App search reliability.

• Resolved issues with collections behavior when filtering assets.

• Improved SEO card rendering and visual consistency.

Performance Improvements

Faster Thumbnail Loading: Improved loading performance for thumbnails across the platform, making browsing quicker and more responsive.

Scenario continues to grow with stronger tools and a broader collection of advanced models across the platform. Each update brings more flexibility and speed, so you can stay focused on creating.

• Image edition: added support for Photoroom models

• Upscale: added support for Magnific upscaler

• Support for previews of rigged models

Bug fixes 🐛:

• Fixed an issue causing large files not to be processed (impacted models: Vidu, Hitem3d, Meshy)

• Search: added support for compliance and governance metadata

• Models: improved input configuration, alignment with workflows

• Workflows: added support for 60+ public apps

Bug fixes 🐛:

• Billing: fixed a rare issue where the wrong price was displayed for OSCU

• Models: improved dynamic input typing allowing for more flexibility

• Handle voxels preview

Bug fixes 🐛:

• Image sequence to video: fixed an issue leading to some frames being inverted

• Image sequence to video: fixed an issue with images with odd width or height

Apps

We have added 56 new Apps to the growing list on your homepage. Each App runs a workflow for you so you can try something in seconds without building a project from scratch.

Here are a few examples of Apps we have recently added:

• Recipe Illustrator
  Turn any dish into a recipe infographic with ingredients, cooking icons, dotted process lines and a final plated image.

• YouTube Thumbnail
  Create bold, clickable thumbnails from images and text with strong framing and overlays.

• Seasonal Asset Reskin
  Give any asset a holiday or event theme while keeping the same art style and structure.

• Cosplay Studio
  Use a person photo and a character reference to generate a realistic cosplay shot in everyday settings.

• Character Expression Sheet
  Produce a clean 3×3 grid of the same character showing different expressions in a consistent art style.

You will find these Apps directly on your homepage. Pick one, upload an image or write a prompt and see the result right away. We are already preparing more Apps to expand the choices available to you.

Read more about Apps in our Knowledge Base.

New Model Thumbnails

All models in Scenario now have new and updated thumbnails.

The goal is simple: browsing should feel clearer and easier on the eyes.

Models now use consistent visual blocks and color categories so you can scan them quickly.

This update covers every model category, including image, editing, video, 3D, textures and audio, which makes the model browser cleaner and much easier to navigate.

The platform is now easier to browse visually and more fun to explore through Apps. You can try ideas in seconds and find models more quickly, which means you can spend more time creating.

More Apps and improvements are already on the way.

Apps Improvements

We are continuing to improve Scenario Apps to help you test and explore capabilities instantly.

You can now share direct App links, making it easier to show teammates or collaborators a transformation, and Featured Apps now appear first in the section to help you discover new options more quickly.

When trying an App, you can now generate multiple variations of the result and view them all directly inside the interface without leaving the page. This makes Apps feel less like a starting point and more like a lightweight creative space of their own.

And as always, this is only the groundwork, with more Apps already on the way.

You can read more about the capabilities of Apps in our Knowledge Base.

New Models and Capabilities

Seedream 4.5 Edit

Seedream 4.5 delivers improved control and higher-quality modifications when transforming an existing image, all while maintaining consistency across edits. It’s a more reliable way to reshape or adjust your visuals without losing the details you care about.

Kling 2.6 I2V and T2V

Kling 2.6 T2V Pro is the newest version of Kling’s text-to-video model. It produces high-quality cinematic clips with smoother motion and improved consistency across frames. The model also supports native audio, including natural voice synthesis, lip-sync, and ambient sound.

Kling 2.6 I2V Pro upgrades image-to-video generation with more natural motion and better lighting. The model turns a single frame into a dynamic 5–10 second sequence with expressive transitions and synchronized audio that matches the atmosphere of the original image.

You can read more about Kling models in our Knowledge Base.

Kling O1 First and Last Frame

Kling O1 generates videos by taking a first frame and a last frame, then synthesizing the motion in between. It creates clean transitions while preserving identity across the sequence.

The model also supports text prompts for guiding style and motion behavior, giving you more control over the look and feel of the final clip.

Models & Tools

Model Tools: A new Model Tools section has been added (behind a feature flag). Models must be configured and published before enabling the tools menu.

Model Tags: Testing and Legacy tags now appear directly on model thumbnails, making it easier to understand model availability and status at a glance.

New Visual Capabilities

Quads Display: A new visualization mode for 3D assets that shows meshes in quads, giving a clearer view of the topology and making geometry easier to inspect.

Vox Generation & Display: Support for Vox generation and visualization has been added, expanding Scenario’s capabilities for volumetric 3D assets.

Bug Fixes

Resolved issues related to caption display during the model training process.

This release pushes Scenario forward on several fronts: faster exploration through Apps, stronger video generation, and clearer tools for 3D work. Each update brings the platform closer to a more powerful and more intuitive creative experience.

We’ll continue focusing on what helps you create with confidence, control, clarity, and speed.

• 3d: new improved mesh converter supporting conversion between various formats: glb, obj, fbx, stl, ply, and vox,

• Models: update compliance and governance metadata

• Apps: add support for 15+ new public apps

• Workflows: add support for unlisted workflows

• Workflows: refine the public route for workflows to help LLM vibe-coding

• Workflows: added support for video and 3d outputs

Bug fixes 🐛:

• Assets: file weight optimization over the Scenario’s CDN

• Assets: optimized CDN dynamic asset transformation for both upsampling and compression

• Workflows: change the default resolution of specific public apps such as “Eight Direction Sprite”, “Hair Studio”, “Logo Placer Pro”, etc.

• Flux Kontext: improved scalability

• Tools: new grid maker

• Gemini: support more than one generation

• Segmentation: use of Meta SAM3

• Video: support for Vidu

Bug fixes 🐛:

• Gemini: fix in ratio determination

• Gemini: better progress report

Apps

We are introducing Scenario Apps, a new way to explore and use Scenario without building a workflow from scratch.

Instead of constructing an entire pipeline, you simply pick an App from your homepage, drop an image or prompt, and see the transformation instantly. The actual workflow steps run behind the scenes.

Apps turn the most common user actions into one-click experiences: swapping weather, jumping through eras, redesigning rooms, cleaning text, posing characters, coloring sketches, creating studio portraits, and more. They help you understand what the platform can do by letting you try it instantly.

If you like the result generated by the App, you can click the three dots in the top corner of the image to convert it to 3D or video, enhance it, or even check the generation parameters.

This launch is only the beginning, making way for a future where you create your own Apps - whether you keep them private or eventually share them - transforming the platform into something built around how you want to work.

New Models and Capabilities

FLUX.2 Models

The FLUX.2 lineup delivers three distinct variants, each designed for different phases of creative work.

FLUX.2 Flex: A versatile, everyday creation model focused on speed and adaptability. Flex generates reliable outputs across a wide range of aspect ratios and resolutions, making it ideal for rapid ideation.

FLUX.2 Dev: This model offers a strong balance between quality and performance. It provides high visual fidelity, making it well-suited for iterative work and experimentation.

FLUX.2 Pro: The flagship FLUX.2 model for production-grade visuals. Pro offers sharper detail, stronger prompt adherence, and premium output quality. This is the model to use when you are producing polished final assets or high-precision visuals.

RetroDiffusion Models

A new family of pixel art models is now available.

RetroDiffusion Tile: A specialized generator for modular, top-down tilemaps. Tiles align perfectly on a grid and preserve terrain coherence, making it ideal for overworlds, RPG maps, dungeons, and other tiled environments that require consistent style and handcrafted pixel charm.

RetroDiffusion Plus: A premium pixel-art model built for high-quality output. RetroDiffusion Plus supports a wide range of pixel aesthetics, from classic 8-bit and 16-bit to modern detailed pixel rendering. It excels at environments, polished character assets and more.

RetroDiffusion Animation
A motion-aware generator designed for sprite sequences. The model supports walk cycles, run cycles, idle poses, and directional animations for both top-down and side-scrolling games. It produces smooth, cohesive character animations and polished VFX sprites tailored for modern and retro-style games alike.

You can read more about the capabilities of RetroDiffusion Models in our Knowledge Base.

Pixel Snapper

Pixel Snapper improves pixel art quality while preserving the original aesthetic. Use it to sharpen edges, stabilize shapes, clean up pixel sprites, and remove noise without losing style.

It is a powerful cleanup companion for RetroDiffusion outputs or any pixel project.

Read more about it here.

Crystal Upscaler Creativity

The Crystal Upscaler now includes a Creativity parameter. Raising Creativity introduces more artistic interpretation and stylistic freedom to your results. Lowering it preserves original structure and fidelity. This allows you to decide how subtle or expressive the upscaled output should be.

Features and Improvements

Model Tags
Models now support testing and deprecated tags. These labels appear directly on model thumbnails so you can understand model availability at a glance.

Gallery Search Memory
The Select from gallery interface now remembers your last search state. Filters and queries persist automatically, improving navigation when browsing large libraries.

GIF Support
You can now upload and use .gif files. This expands compatibility for visual references and source material inside Scenario.

Reference Image Display
The reference panel has been improved when working with a single reference image. The layout is clearer and easier to read, reducing visual clutter while composing generations.

Apps are here, new models are live, creation is faster, and the future drops are going to hit even harder as we move toward node-based automation.

We will keep delivering what matters most: control, quality, and speed, so you can focus on creating.

• 3D: fix an issue in meshy-remesh preventing certain generations

• Models: prepared for the expansion of several new models

• 3D: introduce the experimental Scenario’s voxel generator model (able to take an image as reference)

layer 62/101

• image: new image slicer model

• image: support for the new gemini-3-pro-image-preview model

• image: new reframe model based on gemini-3-pro-image-preview

• image: new upscale model based on gemini-3-pro-image-preview

• image: implement the new Pixel Snapper from Sprite Fusion

  

• image/video: new image sequence to video model

• image/video: new video to image sequence model

Bug fixes 🐛:

• video: increased timeout on the openai-sora2 model

• video: fix a typo in an error message

• Uploads: allow other types of assets to be uploaded with a parentId enabling 3d capture in the frontend

• Models: remove background and reframe as custom models

Bug fixes 🐛:

• Workflows: fixed an issue in the API documentation

New Models and Capabilities

Nano Banana Pro

The new Nano Banana Pro model is now available, offering higher-quality results and more refined control across both text-to-image and image editing workflows. It builds on the strengths of Nano Banana with cleaner details, stronger color handling, and improved prompt alignment for professional-level editing and concept creation.

Nano Banana Pro supports both use cases:

• as a standalone text-to-image model with reference image support

• as an editing model directly inside Edit with Prompts

The model also introduces a series of quality improvements, including 4K-ready output, support for up to 14 reference images, and more stable character consistency across angles and lighting. Text rendered inside images appears clearer, blending between multiple visual sources is smoother, and the model behaves more reliably when handling complex instructions.

You can access Nano Banana Pro from the Image Generation interface or directly through Edit with Prompts.

Hunyuan 3D Part

A new addition to our Hunyuan 3D models, Hunyuan 3D Part generates high-quality 3D parts directly from a single RGB image. The model automatically breaks the object into multiple clean, independent meshes, with no manual segmentation or extra processing needed.

You can read more about Hunyuan 3D models here.

3D Improvements

OBJ File Support

The platform now supports .obj files for 3D models. This makes it easier to export 3D assets and use them directly in common tools such as Blender or game engines.

Bug Fixes and Stability Improvements

Edit with Prompts - Missing Models

Fixed an issue where certain models would disappear from the Edit with Prompts model list. All eligible models now load consistently.

Mini Canvas - Color Selection

Resolved a problem that caused inaccurate or inconsistent color selection in the mini canvas.

Upscale - Cost Estimator

Fixed a calculation issue that affected the cost estimator for the Upscale tool, ensuring accurate credit previews before running an operation.

Model Badges - Thumbnails

Corrected missing or blank thumbnails in model badges, improving visual clarity across the interface.

Safari - Drag and Drop

Fixed drag and drop issues on Safari to improve browser compatibility and ensure consistent behavior across platforms.

• 3d: app.scenario.com now natively supports OBJ, PBR textures and Quads asset generations

• Jobs: you can now get only your workflow execution on GET /jobs (See API documentation)

• Workflows: you can now add a “user approval” node that will require a manual validation before continuing execution

Bug fixes 🐛:

• 3d: fix an issue with mesh properties determination, preventing some meshes from being processed

• 3d: fix an issue that was considering a mesh as an image

• Jobs: fix an issue preventing some jobs from getting accurate progress

• Web Sockets: fix an issue preventing plans with unlimited generations from getting real-time, accurate consumption

• Workflows: fix an issue preventing certain users from executing workflows

• Workflows: various API documentation fixes

• 3D: rendering now supports more native formats without additional conversion

• 3D: add management for PBR textures on tencent-hunyuan-3d

• 3D: various enhancements on the hitem3d model

• video: improve the scalability for openai-sora2 model

• video: improve the scalability and reliability of wan2.2 animate model

Bug fixes 🐛:

• 3D: fix an issue in tencent-hunyuan3d-part preventing the generation of certain meshes

• 3D: several improvements in meshy* models

• video: fix an issue in the progress of the self-hosted wan2.2 animate

New Models

10+ Powerful Qwen Edit LoRAs

Qwen Edit is rapidly becoming one of the most popular open-source editing models, developed by Alibaba and offering powerful, trainable capabilities with a wide range of possibilities. After introducing Qwen Edit as a new model within our “Edit with Prompts” tool, we’ve now added a collection of recently released community LoRAs for Qwen Edit, enabling even more customized editing workflows and finer control.

These LoRAs unlock new creative uses, including product shots, next-scene prediction, object fusion, and more.

For guidance on how to use these LoRAs effectively, be sure to read our dedicated guide: https://help.scenario.com/en/articles/qwen-image-edit-lora-guide

Qwen Camera Control (Qwen Multi-Angle)

We’ve added a dedicated model page for Qwen Edit Multi-Angle, featuring custom slider controls that let you easily generate alternate camera angles of the same product or scene.

Access the model here: https://app.scenario.com/models/model_qwen-edit-multiangle

Read the documentation here: https://help.scenario.com/en/articles/qwen-edit-multiangle-the-essentials

Pixverse Swap

Pixverse Swap is a powerful video editing model that allows you to replace people, objects, or backgrounds within a video while maintaining high visual quality, consistency, and coherence. You can find it in the Video category or directly on its model page (https://app.scenario.com/models/model_pixverse-swap).

As always, don’t hesitate to check out the pinned examples on the model page, and refer to the full documentation here: https://help.scenario.com/en/articles/pixverse-swap-video-element-replacement

Improved UI & Usability

The generation screen has been redesigned to provide a cleaner layout and a more intuitive workflows offering and improved control during creation.

Sketch and Zoom on Reference Images

You can now sketch and zoom directly on reference images from the (“file array component”) within the generation screen, to the left column, to make detailed edits and review closer.

Clickable Model Badges

Model badges are now clickable, taking you directly to each model’s page. They also appear in the “Explore” section, making model identification and navigation faster and clearer.

Smaller UI Improvements

• Update of the 3D Gallery. “3D textures“ have been removed from the preview gallery to improve overall clarity and browsing performance when viewing 3D assets.

• Search Filters. Filters now remain anchored at the top, making it easier to refine results while browsing large sets of assets.

• Low-Resolution Indicator for Training Pairs. Flux Kontext Training now displays a visual indicator for low-resolution training pairs, helping creators quickly spot suboptimal data and maintain higher overall model quality.

• Copy Error Message. You can now copy error messages directly from the notification panel (bell icon). This makes it easier to share exact details when debugging or contacting support.

• LoRA Model Badge: When an image is made with a LoRA, the LoRA badges now appears instead of the base model used, improving visibility and identification.

• Edit with Prompts: Aspect ratios are now preserved when switching between editing models, making it easier to quickly swithc and compare the same type of inference

• Models: you can now access list of all available models on app.scenario.com without being authenticated

• Users: smoother loading of the user’s information

• Workflows: add thumbnails properties to published workflows

• Workflows: new workflows

Bug fixes 🐛:

• Dashboard: new models now relate to the correct category

• Models: you can now list your Flux Kontext Lora generations from the Lora Model’s

• Prompt Agent: all editing models are now working correctly

• Workflows: fix an issue preventing some users from being able to run them

• 3d: automatic image resize (in both dimensions and size in bytes) for tencent-hunyuan-3d

• video: added vidu model

• video: added support for automatic resizing of images in google-veo3* models

• video: support better Google moderation detection for more explicit error messages

• video: extended timeout for google-veo3* models

Bug fixes 🐛:

• 3d: support a queuing mechanism for tencent-hunyuan3d-pro: the 3d model generation demand is pretty high and tencent-hunyuan3d-pro resources are limited, you may experience delays in generation

Tencent’s latest generation of its acclaimed Hunyuan 3D model suite is now available on Scenario as Hunyuan 3D Pro (3.0), with three variants: Single Image Input, Multi-View, and Sketch-to-3D.

It incorporates Tencent’s “Polygen” AI retopology, producing optimized textured meshes for streamlined 3D workflows. Generate high-resolution dense meshes (up to 1.5M faces), or directly produce optimized low-poly meshes with built-in retopology.

You can access the models in your Scenario workspace directly:

• Hunyuan 3D Pro Standard - Single image input, high-fidelity output - direct link (including Optimized Mesh)

• Hunyuan 3D Pro MultiView - Up to 4 reference images (front/back/left/right) for accurate symmetry - direct link (including Optimized Mesh)

• Hunyuan 3D Pro Sketch - Converts hand-drawn sketches to textured 3D models - direct link

As with all Scenario models, you’ll find pinned examples on each model page (see the “Pinned” category in the left menu) that you can preview, compare, download, and reuse.

More documentation on our Knowledge Base (Hunyuan 3D Essentials)

A Game-Changing “Optimized Mesh” Mode (aka PolyGen)

• Automated Retopology - Converts dense meshes to clean, production-ready topology

• 70% Efficiency Gain in asset creation (Tencent internal tests)

• Optimized Mesh Mode - Outputs low-poly, animation-ready assets with quad/tri faces

• Less/No Manual Cleanup Required - Game-ready out of the box

Key Highlights

• Hunyuan 3D Pro uses Tencent’s new hierarchical 3D‑DiT “carving” process that builds a coarse shape and refines details, tripling modeling accuracy over previous versions and supporting up to 1536³ resolution and 3.6 billion voxel outputs.

• Retopology with PolyGen – An Optimized Mesh mode employs Tencent’s PolyGen retopology system to output cleaner, animation‑ready meshes while preserving detail

• Flexible inputs & modes – Generate from one photo, up to four views or even a sketch. Select Standard, Optimized Mesh or Mesh Only modes and set the face count (50 k – 1.5 M) to balance fidelity and file size

• Improved character & texture fidelity – Refinements yield more natural faces, poses and textures that wrap correctly around geometry

• Accurate Texture Mapping - Textures wrap naturally to geometry (fabric, skin, surfaces)

• Occlusion Handling - Predicts hidden geometry when parts are blocked

• PBR Textures - Physically-based rendering for photorealistic surfaces

• Integrated & accessible – Hunyuan 3D Pro is now available in Scenario’s 3D generation tool; it is free to use and comes with Scenario API access (https://docs.scenario.com/)

• 3D: support for tencent-hunyuan3d-pro as leveraging Tencent’s API

Performance improvements 🚀 :

• 3D: 5x factor in 3D mesh preview rendering performance for video previews

Meshy 3D Suite

A complete suite of 3D generation tools is now available on Scenario, enabling fast and flexible asset creation directly within your workspace.

• Text to 3D – Generate detailed 3D models from natural language prompts, with the option to guide results using a style image.

• Image to 3D - Convert any 2D image into a 3D mesh while preserving proportions and visual style.

• Remesh - Optimize 3D models for smoother geometry and balanced topology.

• Retexture - Apply new materials or styles to existing 3D assets while keeping geometry and UVs intact.

All Meshy 3D models are available under the 3D category. Read the essentials.

Qwen Edit Plus

Qwen Edit is now available directly whithin Edit with Prompts. Read the essentials.

New Features and Improvements

Asset Transfer Between Projects

You can now transfer assets directly between projects, simplifying organization and collaboration across teams. See documentation for more info

New Model Page in Topbar

A new Model section has been added to the top navigation bar for quicker access and easier model discovery.

Custom Avatars

Avatars for projects, organizations, and profiles can now be updated directly from the interface. Go to your Organization settings to upload your logo, or your Profile page to set up your avatar.

Flux Kontext LoRA Update

You can now change the type of an uploaded Flux Kontext LoRA from the model page.

• 3d: support for hunyuan3d part as a Scenario managed model

• 3d: add Meshy Remesh & Rigging as a Proxy to Meshy service

• image: enhance scaling performances for openai-image-1, gemini*, flux-kontext models

• image: standardize input parameters for all editing models

• image: enhance scaling performances for Scenario upscale-flux model

• image: add rembg as a Scenario managed model

• video: add support for veo3.1

• video: add support for last frame on wan2.2

Bug fixes 🐛:

• 3d: fix a rendering issue leading to low resolution rendering

• train: improve control image size management

• API Documentation: Categories refreshed for easier navigation and LLM support. Deprecated endpoints removed. Explore here: https://docs.scenario.com

• Assets: Added support for transferring assets between projects while preserving creativity trails (operations that led to asset creation) at the team level.

• Models:

  • Added support for Compliance Metadata.

  • Introduced auto-generated short descriptions (shortDescription) for models. You can also edit these descriptions for your own models. It’s used to improve model search visibility.

• 3d Generation:

  • Added Sparc3d. Starts here: https://app.scenario.com/models/model_sparc3d

  • Improved preview generation pipeline for higher visual quality.

• Edit with prompt: Now supports new editing models dynamically.

• Upscale:

  • Deprecated the POST /generate/upscale endpoint.

  • Use the new POST/generate/custom/{modelId} endpoint with models such as model_scenario-upscale-v3.

  • Learn more in the Upscale Generation section of our documentation.

• Workflows (Private Access): You can now trigger node-based workflows directly via API — with more workflow features coming soon.

• Infrastructure: Rolling out a new serverless GPU infrastructure. Expect faster cold starts and smoother generation performance in the days and weeks ahead.

Sparc 3D Example

Bug fixes 🐛:

• 3D Models: Notifications now correctly trigger websocket updates.

• Various performance optimizations and bug fixes.

Sparc3D is built on a novel sparse representation and an advanced autoencoder architecture that captures fine geometric detail with impressive accuracy - even on open surfaces, thin structures, or intricate parts.

Scenario is the first GenAI platform to integrate Sparc3D, less than 48 hours after the Hitem3D API went live.

Key Features

• Exceptional geometry quality with realistic depth and consistent surface definition

• Two model variants: Standard (general-purpose) and Portrait (faces, busts, and characters)

• Pinned examples available on each model page for preview and download

• Full configuration access including resolution presets and multi-view options

Tips for Best Results

• The “1536*3” mode (Standard or Pro) offers maximum fidelity but may take 5–10 minutes per generation.

• For sharper textures, upscale your input image to 2K or 4K using Scenario Enhance tool before 3D conversion.

• Resulting meshes are dense, making them bases for retopology in Blender or other 3D tools.

Documentation

Documentation and settings details are available here → Sparc3D – The Essentials

• LTX v2 Fast and LTX v2 Pro: the Fast model is designed for rapid ideation and mobile workflows, giving creators instant feedback and real‑time preview while the Pro model delivers high visual fidelity with a fast turnaround, making it ideal for daily content creation and marketing teams:

  • LTX v2 Fast: https://app.scenario.com/models/model_ltx-2-fast

  • LTX v2 Pro: https://app.scenario.com/models/model_ltx-2-pro

• Kling 2.5 I2V Standard is now available on Scenario: create stunning 720p videos with the signature Kling 2.5 quality. It joins the lineup of Pro models already available on the platform: https://app.scenario.com/models/model_kling-v2-5-i2v-turbo-standard

  

✨ Features

• 🖼️ New Cropping Ratios

  • Added additional cropping ratios for greater control: 9:16, 3:4, 2:3, 1:1, 4:3, 3:2, 16:9, and Custom.

    

• 📦 Assets

  • Added model badge and user property display on assets.

    

  • These are enabled by default and can be disabled in the User Profile settings.

    

• 🧠 Model Management

  • Rename Models During Training – You can now rename a model at any point during the training process.

    

  • Bulk Actions for Training – Added bulk tools allowing you to remove backgrounds or delete multiple images in one action.

    

• 💰 Pricing & Compute Planner

  • The Compute Planner now fully includes 3D, Video, and Audio generations in its calculations and estimations.

    

• 📁 Projects

  • You can now deactivate models on individual projects for better control over which models can be used.

    

• 🖼️ Thumbnail Selection

  • You can now select a thumbnail for your Models from your entire asset library.

🐞 Bug Fixes

• Audio Models – Fixed redirection issue: using “Use this model” on audio models now correctly opens the intended page.

• Image Generation – Flux Kontext LoRAs are no longer displayed on the image generation screen.

• Model Zoom – Resolved an issue where zoom scrolling didn’t work properly for pinned images from Model page

• Usage: prepare the endpoint for more advanced metrics such as Active Users or Models’ leaderboard. More to come soon

Bug fixes 🐛:

• Upload: fixed an issue where some meshes couldn’t be uploaded

• Project: fixed an issue that could occur when updating models’ access

• Edit with prompt: fixed an issue with ratio on the Kontext’s model

Two new audio models are now available on Scenario.

• Beatoven Sound Effect: Link.
  Create professional-grade sound effects ranging from animal and vehicle sounds to nature, sci-fi, and abstract tones. This model is ideal for films, games, and digital content that require realistic and diverse sound design.

• Beatoven Music Generation: Link.
  Generate royalty-free instrumental music across genres including electronic, hip hop, indie rock, cinematic, and classical. Perfect for games, films, podcasts, and social content that need consistent and ready-to-use background music.

Both models are available under the Audio category in Create.

Read the essentials.

New Visual Models: REVE Suite (Create, Edit, Remix)

A new set of visual models is now available for greater control and flexibility across image workflows.

• REVE CREATE: Link.
  Generate original visuals from text prompts with strong attention to detail, color, and style. Suitable for concept art, marketing assets, or ideation.

• REVE EDIT: Link.
  Modify or refine existing visuals using natural text instructions. Upload an image and describe the change you want, such as “change the color of the dress” or “replace the background with a sunset beach.” The model performs accurate edits while preserving structure and artistic coherence.

• REVE REMIX: Link.
  Reinterpret and restyle images with creative freedom. Combine or restyle reference images using prompts such as “merge both characters” or “convert to watercolor style.” Designed for experimentation, it enables new aesthetic directions and artistic variations.

New Video Models: Vidu T2V, I2V, and R2V

The Vidu suite introduces next-generation video generation for text, image, and reference-based creation.

• Vidu T2V (Text-to-Video): Link.
  Generate coherent video clips from text prompts with realistic motion, camera movement, and scene transitions. Supports 16:9 and 9:16 formats, with 4-second and 8-second durations.

• Vidu I2V (Image-to-Video): Link.
  Transform static images into dynamic videos with subtle motion, character movement, and environmental animation. Optionally refine with descriptive prompts for style or mood control.

• Vidu R2V (Reference-to-Video): Link.
  Upload up to seven reference images such as characters, props, or environments and generate cohesive videos guided by a text prompt. The model preserves visual identity and style consistency across frames, enabling professional-grade storytelling and branded content creation.

All Vidu models are available under the Video category and are compatible with Prompt Spark for better prompt optimization.

Read the essentials.

New Upscaling Model: Crystal Upscaler

Crystal Upscaler is a high-precision image upscaling model designed to enlarge and enhance visuals with minimal quality loss.
With a single Scale Factor setting ranging from 1× to 16×, users can increase image size while preserving detail, texture, and sharpness. Crystal Upscaler intelligently reconstructs features to produce clean and refined results even at extreme magnifications.
It is ideal for preparing assets for print, restoring older visuals, or upgrading artworks and designs for high-resolution display.

Access it here.

Read the essentials.

Documentation & Support

As always, our updates come with documentation and best practices to help you explore new models and workflows quickly.
Visit the Knowledge Base for more detailed guides and examples.

Veo 3.1 and Veo 3.1 Fast are now available on Scenario!

• Veo 3.1 (for Pro plans and above): link

• Veo 3.1 Fast (for Pro plans and above): link

Veo 3.1 delivers cinematic text-to-video and image-to-video generation with greater clarity, improved subject stability, and enhanced lighting realism.

Compared to Veo 3, this new version introduces first and last frame control, allowing creators to define both the opening and closing moments of a shot for smooth, coherent transitions. It also adds multi-reference guidance, enabling the use of multiple reference images to direct the style, subjects, and environments of your video for finer visual control.

As of Oct 2025, Veo 3.1 stands as the state-of-the-art model in text-and-image-to-video generation.

As always, you’ll find pinned examples on each model page. For stronger video prompts, use Prompt Spark to generate or refine cinematic instructions that make the most of Veo’s expanded aesthetic range. More information is available in our Knowledge base.

Gallery & Generation

Image Comparison in Gallery

You can now compare two images directly in the gallery. Select any two images and click “Compare Images” for a focused comparison view (Split or Side-by-Side).

Aspect Ratio in Gemini 2.5 (aka Nano-Banana)

The Gemini 2.5 editing model (Nano-Banana 🍌) now includes an Aspect Ratio setting, offering finer creative control and consistent compositions across generations. Besides “Auto”, the Available aspect ratios are :
21:9, 16:9, 5:4, 4:3, 3:2, 1:1, 2:3, 3:4, 4:5, 9:16

Access Gemini 2.5 it in the “Edit with Prompts” tool.

“Select from Library“ Optimization

The “Select from Library“ feature has been optimized for faster performance. Generated content now appears instantly when no filters or queries are applied, for a smoother workflow and near-instant asset refresh.

Additionally, recently deleted assets (previously displayed in low resolution) will no longer appear.

Projects & Collaboration

We’ve streamlined navigation to better understand how “Members” participate in various “Projects”:

1. You can jump directly from the “Projects list” to the corresponding members involved in that specific project.

   

2. From the list of members list, you can open a dashboard link pre-filled with both author and project filters, for faster visualization.

   

3. When viewing your organization’s member list, you can now hover over the “i” icon to see which projects each user belongs to.

   

⚠️ “Solo” vs. “Collaborative” Projects

You can now create and manage “Solo Projects”, designed for creators who need privacy within a shared Project. In a Solo Project, each member can see, generate, and manage only their own content & assets (including images, videos, and models) - while other members in the same project cannot view/access or edit that creator’s content.

This configuration is good for:

• Confidential production workflows, such as unannounced game assets, unreleased concept art, or client-specific visuals.

• Individual R&D or experimentation, where creators prefer to test ideas independently before sharing them with the team.

• Enterprise teams that require strict content isolation for security or compliance reasons.

🚨 Solo Projects are available exclusively for Enterprise plans.

By contrast, “Collaborative“ Projects continue to function as before: all invited members share a common workspace where every generated asset, model, and training is visible and accessible to all participants.

You can switch between Solo and Collaborative modes at any time, allowing teams to adapt as project requirements evolve — for example, starting with a Solo setup for early concepting and later converting it into a Collaborative Project for production or review.

Enterprise Pricing Page

A new Enterprise Pricing page is now live, offering transparent details for large-scale teams and organizations. Check it here.

Training & Models

Draft Model Deletion

You can now delete draft models directly from the Training list

Training Page Updates

We’ve implemented various UX refinements and fixes across the Training and Model pages for improved clarity and smoother performance. New options include quick shortcuts to tutorial videos and a refreshed “Sort By“ menu (Newest / Oldest / Name (A–Z) / Name (Z–A)”

UI & Experience

Toasts Position & Duration

Success and Info toasts (notifications) now appear in the bottom-right corner and remain visible for half the previous duration, creating a cleaner rhythm and keeping focus on your workspace.

As always, our updates come with documentation and best practices to help you explore new features and models quickly. Check out the Knowledge Base for more detailed workflows and guides

• Model: Support for Veo 3.1 and Veo 3.1 fast with reference image and last frame. t2v, i2v, flf

• Pricing: More complete mechanism to estimate any api call cost

• Pricing: More details are now available when a user pauses subscription

• Asset Data Management (ADM): You can now copy an asset to another project

• Prompt Agent: Added enhanced support for gemini-2.5 to always get the best prompt

Bug fixes 🐛:

• Pricing: fixed an issue that may have prevented a user to access the billing portal

• Model: fixed an issue preventing to pin images on Kontext fine tunes

• Model: fixed an issue that may caused incorrect caption to be used in trainings

• Reframe: fixed an issue preventing to extend too large images

• Search: fixed an issue that was causing some deleted asset to show up in search results

• Video: Add support of OVI t2v and i2v video models

• Models: Enterprise admins can now prevent the access to models at projects level

• Models: Added Krea model training

• Project: You can now share resources in a project set to “Solo mode 🥸”. As a creator you will be able to work without others to see your experimentations

Bug fixes 🐛:

• Models: several bug fixes and improvements on Kontext training

We’re proud to introduce one of our most significant updates yet, and another major step forward for personalized AI models, giving Scenario users even more control over their output and, in turn, higher productivity.

Flux Kontext LoRAs bring a new way to train and create editing models in Scenario: think of it as “training your own nanobanana model, built from and for your own art or brand.”

Until now, training a LoRA model was mostly about teaching AI how to reproduce a style or a character. With Flux Kontext LoRAs, you go further by also teaching AI to understand and follow your own editing instructions: how to adjust lighting, change poses, apply stylistic transformations, or refine any visual detail you can demonstrate through paired examples.

Flux Kontext LoRAs are trained using before-and-after image pairs, along with a short caption describing the change. For example, “Relight this image as golden hour” or “Turn this character into a LEGO figurine.” The model learns not just the visual result, but the intent behind your transformation, making your edits faster, more consistent, and fully aligned with your creative direction.

Once trained in Scenario, your Flux Kontext LoRA becomes a model you can load directly in the “Edit with Prompts” interface. Simply load an image, add your LoRA, and write your instruction. Your Flux Kontext LoRA delivers the transformation accurately, consistent with the editing logic it was trained on.

Complete documentation is available on our Knowledge Base and YouTube channel. Learn more:

• How to train a Flux Kontext LoRA

• How to build your training dataset for Flux Kontext LoRAs

• Flux Kontext LoRA use cases

Flux Kontext LoRA training is only available for “Max“ / “Enterprise“ and selected legacy plans.

Redesigned Training Interface, with Support for Epochs

We’ve completely redesigned the training interface to make the entire process smoother and more intuitive. The new workflow simplifies every step, from selecting the base model and uploading your dataset to adding captions, setting parameters or test prompts.

During training, Scenario now automatically generates preview outputs using your test prompts (or test pairs), so you can monitor how your model is learning, in real time. You can easily compare results across epochs, evaluate outputs side by side, and select the version that delivers the best output for your goals.

New Video Model: Open AI Sora 2 🚀

We’re excited to introduce OpenAI’s latest video model, Sora 2, now available in Sora 2 and Sora 2 Pro versions on Scenario. Both support text-to-video and image-to-video generation, allowing you to create clips of 4, 8, or 12 seconds in 16:9 or 9:16 formats, at 1792 px or 1280 px resolution.

As always, you’ll find curated examples on each model page, complete with prompts and settings you can reload and reuse instantly.

• Sora 2 (available for all paid plans): link

• Sora 2 Pro (for Pro plans and above): link

We’re also releasing these models with our Prompt Spark assistant - it will automatically generates prompts specifically optimized for Sora 2, per OpenAI's official recommendations and documentation.

• If you upload a start frame, Prompt Spark creates a relevant prompt based on it.

• You can also add a few keywords and click "Rewrite“ to instantly refine your prompt... no need for prompt-engineering deep dives.

Stay tuned for more documentation in the Knowledge Base with additional examples or creative use cases surfaced by the user community (spatial prompting, collaging multiple input frames, and more)

Churnkey Integration

We’ve integrated Churnkey into Scenario. Users now have the ability to pause their subscription as an alternative to full cancellation, making account management more flexible.

Public Pages / Resources

Support Page Redesign - The Support page has been completely redesigned and is now publicly accessible, making it easier for anyone to find documentation, guides, and troubleshooting resources without signing in.

New “Enterprise/Support” Pages - A new How Do We Sell page is now live, for more transparency on Scenario’s offerings and plans.

Miscellaneous

Negative Prompts on Skybox - Skybox generation now supports negative prompts, allowing you to exclude elements from a scene and achieve cleaner, more controlled outputs.

Upcoming Fixes & Known Issues (Flux Kontext LoRAs)

The following issues will be addressed in upcoming release(s):

• Retraining Kontext LoRAs - not functional yet

• Pinning images for Kontext LoRA - not functional yet

• Percentage display - may appear inconsistent depending on the screen or device used

• Duplicate as Model - feature planned to allow creating multiple versions of the same model with different epochs in the model list

• Epoch images in Training Model list - not displayed yet

• Kontext image epochs - sometimes display incorrectly due to improper aspect ratios

• Video: Support for Sora 2 in the API

• Prompt Agent: Custom Prompt generation tailored to the Sora 2 and Sora 2 Pro requirements. You will always get the best prompt based on your ideas and context

Bug fixes 🐛:

• Cleaner error message in case of third party moderation (Google Veo3, Sora 2 OpenAI, etc.)

Lipsync: Veed Fabric 1.0

Editing: Qwen Image Edit Plus

Video: Wan 2.2 Models

Wan 2.2 Animate (“Replace” & “Move)

1. Wan 2.2 Animate Move: Animates a static character image to follow the motion and expressions from the video, generating an original animation with a newly generated background.

2. Wan 2.2 Animate Replace: Replaces a person in the video with the character image while preserving the original video's lighting, environment, and performance.

Explore these powerful animation tools further in the Knowledge Base.

Wan 2.2 Reframe & Wan 2.2 Outpainting

1. Wan 2.2 Reframe: Intelligently adjusts video composition when changing aspect ratios, going beyond simple cropping to ensure key subjects remain framed. Ideal for optimizing content for various platforms.

2. Wan 2.2 Outpainting: Expands videos in any direction, creating new, contextually appropriate content that naturally extends the scene while maintaining lighting, perspective, and environmental consistency.

Video Editing: Lucy Edit Models

Video: Seedance 1 Pro now with Last Frame Image

3D: Rodin Hyper3D Gen-2

• Billing: preparer backend to pause subscriptions

• SSO: self onboarding

Bug fixes 🐛:

• Models: fix an issue preventing the update of a thumbnail when the thumbnail had been deleted

New Models Added

1. OmniHuman 1.5 (Bytedance) - A leading lipsync model and the successor to OmniHuman 1.0. Upload one image and one audio (generated on Scenario if needed) and animate in one click. Easy to use and powerful. Direct Link

2. Kling 2.5 Turbo Pro - Major upgrade in the Kling model series, now 2x faster and 20% cheaper than Kling 2.1 Pro. Delivers cinematic motion with higher quality. Available in Image-to-Video (I2V) and Text-to-Video (T2V)

3. Wan 2.5 (Alibaba) - Wan 2.5 is a serious Veo 3 competitor, able to generate with audio at 1080p resolution. Available in Image-to-Video and Text-to-Video.

As always, we do not just release a model without documentation or support. You will find dozens of relevant examples in the “pinned“ section of each model page. More importantly, you can use Prompt Spark to automatically write or rewrite strong video prompts. And for more details, our Knowledge Base has been updated

New Features & Capabilities

Seedream Edit

Following the release of Seedream 4.0 with the Image Models (last week), we are also introducing Seedream Edit, available in the “Edit with Prompt” section. Seedream Edit is one of the most powerful editing models available, joining GPT, Kontext, Gemini 2.5, and Runway Gen4. Qwen Edit will be added soon.

Note: “Sequence Mode” is not available in Edit with Prompts as this stage. Use the the image model here for accessing Sequence Mode.

“Use as Reference” for Videos

You can now take any generated video and hit “Use as Reference” to use it as input for future generations. The default model (pre-loaded) is currently Luma Reframe (for format changes and outpainting). Change to other options like MMaudio (add sound to the video), Lipsync models, Runway Aleph (instruction-based editing), and more!

Churkney (Behind Feature Flag)

An experimental “Churkney” feature is now available for selected users under a feature flag, as we are testing retention offers.

Plan Improvements

• Subscription Modal Update – When subscribing to a plan, users can now choose credits and seats directly in the modal for a smoother checkout flow.

• New Limits – Updated seats, projects, and model limits for Starter, Pro, and Max tiers to improve scalability and clarity.

• Legacy Pro Plans – Users on legacy Pro plans now have full access to the dashboard.

User Management

• Member Search – You can now search for users in the members list, making team management easier, especially for larger teams. More user management features are planned.

Bug Fixes

• Sketch on Large Images – Fixed an issue that caused sketching on large images to fail.

• Upscale Gallery Selection – Fixed an issue where “Select from Gallery” only showed images when using Skybox or Texture Upscaler.

• Billing: Updated limits for models and projects

• Billing: models now expose which criteria enter in Compute Units calculation

• Image Generation: Added support for negative prompts with Flux-based image generation.

• Models: Transferring models notification got more actionable with Model’s name and links

• Models: Added support for training image pairs and support for Kontext Training

• Explore: Improved model discovery and filtering

Bug fixes 🐛:

• Billing: Some notifications were missed

• Billing: An issue may occurred for some upgrade. It is now fixed

• Models: Clearer messages and improved model transfer error transfer code

• Models: Changed backend to improve reliability and scalability

• Edit with prompt: support Seedream 4

• Models: better classification of uploaded models, we maximized the success rate

• Offboarding flow: prepare for easier user experience

Bug fixes 🐛:

• Various bug fixes and improvements

• Edit with prompt: Flux Kontext edition now supports LoRA

• Edit with prompt: major leap in multi-reference image edition with Flux Kontext

• Search: a deleted asset will now be removed faster from your search results

• Elevenlabs Sound Effect v2: disable loop by default

Bug fixes 🐛:

• GET /models will now match your type filter

• Edit with prompt: in certain circumstances, input validation was incorrect

• Enterprise: non admin at the organization level can now create API Keys at the projects level

• Other bugs fixes and performance improvements

Kling AI Avatar Pro: Lifelike Talking Avatars

1. Access the Model

2. Read the essentials

MM Audio: Video-to-Audio Synthesis

1. Access the Model

2. Read the essentials

MinMax Music 1.5: AI Song Generation with Vocals

1. Access the Model

2. Read the essentials

What’s new

• Up to 10 reference images for richer context

• Batch generation of up to 15 consistent outputs in one run

• Advanced editing tools: combine characters, vary scenes, remove objects, adjust poses & expressions, retouch styles, and more

• Affordable pricing: only 5–6 credits per image

We’ve also prepared documentation with 20+ examples to help you get started: Seedream 4.0 – The Essentials.

👉 Try it now: Seedream 4.0 in Scenario (Direct link for now, will appear under "Edit with Prompts" soon.)

• Asset Data Management: generated and uploaded audio files now automatically include

  • A title

  • An audio description

  • A transcript for dialogues

• Audio: more appealing and relevant preview generation

• Enterprise Tier: as an admin of an organization, you can now enable and disable models

• Updated Google Veo3 and Veo3 fast

  • Negative Prompt to further control your generations

  • More 9:16 Aspect Ratio to create portrait videos

  • Flexible duration 4, 6 and 8 seconds

  • Massive Price Reduction

• Audio API

  • ElevenLabs TTS Multilingual v2

  • ElevenLabs TTS Turbo v2.5

  • ElevenLabs TTS v3 for higher quality and flexibility

  • New Audio To Audio capability

• Recommendations API

  • Always get the most appropriate models’ suggestions

Bug fixes 🐛:

• Improved message reliability when generating media with the wrong parameters

• Google Gemini became more reliable and more flexible. You can generate more images, faster

We’ve expanded Scenario with AI-powered audio generation — giving you a complete creative workflow for sound, speech, and animation, all inside Scenario.

Text-to-Speech 🗣️

Turn any text into natural, expressive voiceovers.

• Models available: ElevenLabs Multilingual, Resemble AI Chatterbox (more coming soon).

• Control voice ID, stability, speed, and other parameters.

• Create clean, production-ready narration or dialogue in seconds.

Text-to-Music 🎵

Generate original music by simply describing mood, genre, or style.

• Models available: Meta MusicGen, Google Lyria 2.

• Supports everything from ambient backgrounds to cinematic scores.

• Example: “A melodic adventure track blending airy synths and soft plucks.”

Text-to-Sound Effects (SFX) 🔊

Design polished, loopable SFX with a short text prompt.

• Model available: ElevenLabs SFX.

• Perfect for ambient environments, UI sounds, and game effects.

• Example: “Campfire crackling” or “Glass breaking underwater.”

Lip-Sync Animation 👄

Bring avatars and characters to life with synchronized speech.

• Available in the Video interface under Lip Sync models.

• Models available: OmniHuman (realistic avatars), Pixverse (stylized motion), Sync (redubbing videos), Kling (short clips).

• Upload an image or video, add or generate audio, and Scenario syncs the mouth (and gestures in some cases) to the voice.

• New Pricing Plans – Introduced fresh pricing tiers with a redesigned layout, making plan selection clearer and more flexible.

• Visit https://app.scenario.com/team?tab=plans for more.

Image Generation & Editing

• Nano Banana Models (Gemini 2.5 Flash Image) – Added the Nano Banana model to the Edit with Prompt workflow, along with the Nano Banana T2I image generation model.

• Adjusted Default Ratio – The default aspect ratio for image generation is now 4:3.

Video Generation

• Video Comparison Slider – Added a slider to compare video2video outputs side-by-side.

• Pixverse V5 Model – Now available for video generation, offering smoother motion, sharper visuals, smarter prompt control, and both first & last frame support.

• Seedance 1 Pro Default – Seedance 1 Pro is now the default video model.

• .mov Upload Support – Users can now upload .mov videos without compatibility issues.

Pixverse v5: a big upgrade for the Pixverse lineup

Pixverse V5 expands the creative toolkit with fast, reliable AI video generation. Building on v4.5, it delivers smoother motion and ultra-sharp visuals, smarter prompt control with negative prompts, plus first AND last frame support.

Model Details:

• Modes: text-to-video, image-to-video, support first frame, last frame, and negative prompts

• Resolution & Ratios: 360p, 540p, 720p, 1080p | 16:9, 4:3, 1:1, 3:4, 9:16

• Duration & FPS: 5s or 8s videos

• Styling: Variety of effects and motion modes for creative control

Try it now

Pixverse V5 is available now on Scenario. You can try it here: app.scenario.com/es/models/model_pixverse-v5 or visit the Pixverse documentation

UX & Interaction

• Tagging Shortcuts – Introduced keyboard shortcuts to streamline the tagging process.

Smart, fast, and reliable.

Nano Banana (aka Gemini 2.5 Flash Image) is the state-of-the-art text-to-image model by Google. It’s currently in two versions on Scenario: one for editing images (Nano Banana) and the other one for generating images (Nano Banana T2i)

Try Nano Banana Edit

Nano Banana Edit is joining GPT, Kontext and Gen-4 within "“Edit with Prompts” (https://app.scenario.com/edit-with-prompts). Simply open an image, choose Edit with Prompts → Nano-Banana 🍌, and type your instructions. You can:

• Edit with plain text - no training, no manual tweaks

• Lock consistency - across characters, objects, and styles

• Use up to 6 reference images - for more visual context

• Get prompt assistance (Prompt Spark) - for sharper edits

• Reuse parameters in one click - for rapid iterations

Use it to add, remove, or modify objects. Change backgrounds and apply artistic styles. Transfer styles instantly (oil painting, sketch, cartoon, and more). Compose full scenes with multiple characters and elements …and more!

Try Nano Banana T2i (Text-to-Image)

Generate images with Nano Banana T2i, from text prompts – Visit Nano Banana – T2i or access the model via the Create tab → Images and select it in the gallery. This model excels is handling diverse styles, complex scenes, and especially rendering text within images, with high accuracy.

• Reframe (Expand) Videos - Luma Reframe allows you to expand your existing videos beyond their original framing - perfect for reframing content for new aspect ratios, adding context to a scene, or creatively extending your footage without reshooting. Try it now or learn more on our Knowledge Base

• Topaz Image Upscale - a high-quality, detail-preserving upscale option - joining our existing upscaling tools (Scenario Style-Preserving Upscale, Recraft Upscale and Luma Upscale)

New Free Plan

• We’re introducing a free tier with 50 daily credits, giving more users the opportunity to explore Scenario’s capabilities. This is a first test—GPU usage is costly, so we’ll be monitoring activity to prevent abuse and ensure a fair experience for everyone. Based on our findings, we may expand availability or introduce restrictions as needed.

Asset & Generation Improvements

• Upscaler Prompt Import – When sending an asset to the Upscaler, the original prompt now imports automatically for a smoother workflow.

• Default Model Updates

  • Video Upscale now defaults to Topaz.

  • Image Generation now defaults to Imagen 4 Ultra.

• Convert to Video UX – The reference image and prompt sections now open automatically (when applicable) for faster setup.

• Convert Section Reorder – “Convert” (to Video or 3D) now appears above “Edit” in dropdown menus for easier access.

• Upscale Tutorial – Added a dedicated video guide on the Upscale page to walk you through the process.

Dashboard & Tracking

• Team Usage Cleanup – Removed team members now display correctly on the usage dashboard (no more “unknown” labels).

Bug Fixes

• Asset Selection Uploads – Fixed a bug that could block asset uploads.

• Fullscreen Video Autoplay (Mobile) – Stopped videos from auto-playing in fullscreen mode while browsing the gallery on mobile.

• Portrait Video Cropping – Fixed incorrect cropping for portrait videos in fullscreen mode.

Other Changes

• Removed “Cancel Plan” Button – The cancel option has been removed from plan management.

• Vector Search Failure Tracking – Improved analytics to detect repeated vector search failures more reliably.

1. Runway Aleph – Advanced in‑context video editing

2. Ideogram Character – Single‑image character consistency model

3. Qwen (sharp text rendering)

4. Dreamina (Seedance 3 variant)

Runway Aleph (Video)

Runway Aleph takes video editing to the next level. Unlike generative models that create new clips, Aleph is built for in-context editing of existing footage. It analyzes scene context—lighting, spatial layout, motion—and enables seamless modifications via simple text prompts. More details on https://help.scenario.com/en/articles/runway-aleph-the-essentials

Some capabilities include:

• Generating new camera angles (e.g., reverse shots, aerial views)

• Predicting the next shot in a sequence

• Relighting scenes—complete with coherent shadows and color temperature

• Adding or removing objects and subjects (e.g., crowds, props)

• Re‑texturing or swapping elements in the scene

• De‑aging or altering a subject’s appearance

• Creating green‑screen style extractions

• And more!

Ideogram Character

Ideogram Character lets you create consistent character variations from a single reference image—no multi-image training required. Once you define your character, you can reproduce them across poses, styles, scenes, and lighting setups, effortlessly. More details on https://help.scenario.com/en/articles/ideogram-character-single-image-character-consistency

Qwen & Dreamina 3.1

Qwen-Image is a diffusion model by Alibaba Cloud (Aug 2025) known for industry-leading text rendering. Its architecture embeds text across the image grid, producing highly readable English and Chinese text in complex scenes. It handles intricate, multi-element prompts with consistent style, making it ideal for signage, documents, logos, multilingual designs, and other text-heavy visuals.

Dreamina 3.1, from ByteDance’s CapCut team, uses a 5-element prompt structure—subject, description, style, context, narrative—for predictable, high-quality results. It excels at water scenes and reflections, and offers diverse, artistic styles with strong prompt adherence. Perfect for landscapes, portraits, and any project needing rich narrative context.

Our new Scenario Upscale model is now live! 🎉 You can now use any LoRAs, including your own Custom-Trained models, to achieve more consistent upscales with no style drift, scaling up to 8x.

This update ensures that your upscales maintain the original style and details, without any unwanted changes, resulting in clearer and more accurate visuals. You can try it on https://app.scenario.com/enhance. More technical details are available on our Knowledge Base: https://help.scenario.com/en/articles/enhance-upscale

Scenario Fast Upscaler

In addition to the new model, we've kept the previous upscaler, now renamed Scenario Fast Upscaler. It offers faster and more affordable upscaling, but with less control and consistency compared to the new model.

Scenario Skybox Upscale & Scenario Texture Upscale

Both Scenario Skybox Upscale and Scenario Texture Upscale have been upgraded and are now available in the model section or via any texture/skybox input.

Recraft Upscaling Models

We’ve also released two Recraft Upscaling Models (“Crisp” and “Creative“ options). Expect more image upscaling models to be added in the future.

Video Upscale: Runway & Topaz

We’ve also introduced Runway Video Upscale (link), which is joining Topaz-powered Video Upscale (link).

Improvements

Bulk Download: The download modal now displays correct file extensions and options depending on whether you're downloading videos or 3D models.

“Retrain Model“ Behavior Updated: When retraining a model, the system now only copies the training images - not the parameters - to the new model, ensuring a cleaner and more predictable setup.

Bug Fixes

Edit with Prompts – Ref Image Display: Fixed an issue where reference images were not visible in the asset detail view for edit-with-prompts assets.

Large Image Uploads: Resolved a bug that could block large image uploads under certain conditions.

Video Search Stability: Addressed a crash occurring when users searched for videos.

New LoRA Trainer 🚀

We've rolled out a major upgrade to our LoRA trainer, now featuring more advanced parameters like Epochs, Batch Size, and Repeats. This new version delivers significantly higher-quality and more consistent outputs - even with default settings, thanks to our built-in automated captions and automated prompting (Prompt Spark).

Try it now at https://app.scenario.com/models/new, or retrain any of your existing models via the “Retrain” option in the 3-dot menu (to the top of the Model Page). Get more details on the new parameters on the Scenario Knowledge Base.

⚠️ For now, you can only access and download the last epoch (e.g., if you train for 10 epochs, you'll get epoch 10). Multi-epoch comparison and download are coming soon!

Restyle Tool Update

A new version of the “Restyle” tool is available - updated for 10x better quality and more consistent output. Try it on https://app.scenario.com/restyle. Learn more about Restyle on the Scenario Knowledge Base

Use Restyle as an powerful option for Sketch-to-Render. It's similar to ControlNet + Sketch Mode when using a LoRA, however no training is required. Upload one Structure Image and up to 10 Style Images, and generate “zero-shot” results - fast, easy, and effective.

Improvements

Homepage – “Recent Creations”

You can now view your latest generations directly on the homepage for quick and easy access.

3D Viewer – New Shortcuts

We've added handy keyboard shortcuts to speed up navigation and preview in the 3D viewer:

• Press S to close the left panel

• Press X to toggle Exploded View (when Generating in Parts / cf. Partcrafter)

• Press W to toggle Wireframe View

Compute Units Indicator

The “Generate“ button will turns orange when your action exceeds 400 Compute Units, helping users better manage resource usage.

Video Generation – Sketch on Reference Images

When using a reference image in video generation, you can now sketch over it - great for adding spatial prompts or visual annotations.

Canvas Tagging

You can now add tags to your canvas projects to better organize and find them later.

Sticky “Explore“ Filters

Filters in the Explore section now stay active when navigating back from an asset page - no need to reapply them.

Team Member Count

The total number of team members is now visible directly in the member list.

Cost Calculator - Editable Estimated Number

The “Estimated Number” field is now fully editable, letting you input and adjust values directly.

Skybox Default Style

The default skybox style is now set to “Cinematic” for a richer, more polished look out of the box.

Skybox “Vary” Button

We’ve added a “Vary” button to skyboxes, making it easy to explore alternative visual variants with a single click.

Scenario Bot – New Questions

The Ask AI bot now includes two helpful new prompts: “What are Scenario's main workflows?” and “What models are available?”

Bug Fixes

Model Filter Visibility - Fixed an issue where certain models were not appearing in filters, even if they had been used previously.

Recreate Button / Infinite Loading - Resolved a bug that caused infinite loading when clicking the Recreate button on some models.

3D Viewer Glitches - Addressed a series of minor display and interaction glitches in the 3D viewer for a smoother experience.

Organization Creation & Deletion Rules - We’ve tightened up org management logic with the following rules: (i) You can’t delete your last remaining organization and (ii) You must be on a paid subscription with at least one existing organization to create a new one

Right-Click Support on Images - You can now properly right-click → open in new tab on images across the platform.

At the core of this update is PartCrafter, the first generative model capable of synthesizing multiple 3D components from a single image. Unlike existing “single-block” AI mesh generators, PartCrafter intelligently separates your input object into defined components it can recognize (such as arms, wheels, panels, etc.). These parts are cleanly segmented, each with its own geometry and structure.

You can adjust the number of parts and other inference settings, and download as GLB for further editing, animation, or 3D printing.

Test the model now! You can review pinned examples showing its capabilities on various input images.

Make sure to check our knowledge base to learn more tips, best practices, and examples: PartCrafter: The Essentials

Image Generation: Minimax Image 01

Minimax Image 01 is a text-to-image and image-to-image model. While it is capable of generating in multiple visual styles, the model is especially strong in realism, delivering realistic lighting, shadows, and textures for both characters and objects. Users can provide a reference image to generate the same character in different scenes and outfits with consistent appearance. As an image-to-image model, it can also transform or enhance images based on new descriptions or stylistic instructions.

Video Generation: Hailuo 2

Hailuo 2 is Minimax’s latest text-to-video (T2V) and image-to-video (I2V) model.

It’s widely praised for its exceptional visual realism, offering highly detailed simulations of motion, physics, and textures — even in scenes featuring dynamic lighting or complex elements (like water, fire…). Prompt responsiveness stands out, with the model interpreting nuanced prompts with precision. Creatives will appreciate its built-in cinematic camera moves, including pans and zooms. It also delivers consistent frame coherence and strong character persistence, even during action-heavy sequences.

As of June 2025, Hailuo 2 ranks among the top performers in global video benchmarks and is often cited as a serious rival to Google’s Veo 3. It delivers state-of-the-art quality with record-breaking cost efficiency, making it a reliable and highly capable AI video generator with outstanding cinematic control.

In addition to the Hailuo 2 release, we’ve also enabled negative prompts for the following video models: Kling v1.6 Pro, Kling v1.6 Standard, Kling v2.0, Kling v2.1 Master, Lightricks LTX Video 0.97, Lightricks LTX Video, Pixverse v.4.5, Pixverse v.4 and VEO3 Fast.

New Model variant: GPT-Image with High/Low Input Fideliy

We’re introducing GPT-Image-1 with the High/Low Input Fidelity options, for improved image generation and prompt-based editing.

The model now offers two modes, with High mode providing superior accuracy in preserving overall style and important details (like faces, text, logos…) during prompt-based edits - at additional compute cost. This is valuable for professional workflows involving portrait editing, brand asset modification, and scenarios where visual consistency is paramount.

The parameter is available in the bottom left column in settings when selecting GPT-Image on Edit with prompts (Default: Low)

New Features

• A new “All” tab is now available on Home and Explore, allowing you to browse a mixed feed of videos and images.

• Added a “Reuse Image” option for 3D assets (accessible via the 3-dot menu). This will reload the original input image directly into the inference panel on the left.

• In Edit with Prompts, for models that require a plain background, the “Remove Background” button is now replaced with “Set White Background”, which yields better results than using a transparent background (alpha channel).

• On custom models, the Negative Prompts and Optional Images sections are now collapsed by default for a cleaner UI.

• Prompt Spark now uses Contextual Mode across all interfaces except in Edit with Prompts.

• The Recent Tasks view now displays error messages associated with failed jobs, if any.

Improvements & Fixes

• Fixed an issue where the column count slider in model and canvas search was not functioning properly.

• Resolved a crash on Canvas (Retouch) that occurred when performing multiple rapid undelete/delete actions.

• When launching “Retouch” on an image created by an Editing Model or custom Base Model (e.g., Gemini, Flux Kontext, Runway Gen4 etc), Flux Dev now loads by default (since editing models are incompatible with Retouch)

• Corrected the misplaced tab indicator on the navigation bar during app load.

• When importing an asset from Edit with Prompts for upscaling, reframing, or restyling: the original prompt no longer loads automatically.

We’re excited to roll out over 15 new image and video models in this release, giving you powerful new tools to generate and edit your content with precision and speed. Below are some standout models you’ll definitely want to explore:

Image Editing with Runway Gen-4

We’ve just upgraded the “Edit with Prompts” toolkit by adding Runway Gen-4, joining leading editing models like Kontext, GPT‑Image, and Gemini for unmatched editing precision.

Gen-4 excels at keeping styles sharp and subjects consistent across iterations, with support for up to three input images. Effortlessly mix subjects, outfits, or backgrounds to create high-quality edits with unprecedented speed and accuracy.

Image Generation with Imagen, Ideogram & more

You can now generate visuals with the most advanced image models, including Google’s Imagen 3 & 4, Ideogram v3, Bytedance’s Seedream 3, Luma Photon, or Recraft v3. Each model brings unique strengths. Together, they offer a mix of style presets, mask-based control, reference image support, prompt optimization, or flexible resolutions. More details below.

Video Generation with Veo3 Fast

We’ve also added Veo3 Fast—Google’s state-of-the-art video model. It delivers the same high-quality output and strong prompt adherence as the original Veo 3, with native audio and faster, more cost-efficient generation.

Check out these unique features with the new image models!

New models: Ideogram V3 Quality, Ideogram V3 Balanced, Ideogram V3 Turbo, Recraft V3, Recraft V3 SVG, Luma Photon Flash, Luma Photon, Bytedance Seedream 3, Imagen 4 Ultra, Imagen 4 Fast, Imagen 4, Imagen 3 Fast, Imagen 3.

Imagen Family
Google's Imagen 3 and 4 are available in Fast, Standard, and Ultra tiers, each offering increasing levels of quality and precision. Imagen 4 supports resolutions up to 1408×768 and produces outputs with detailed textures, accurate fabric rendering, and strong text layout capabilities for mixed-media content.

Ideogram Family
The Ideogram Family offers three modes—Turbo, Balanced, and Quality—designed for different performance needs. It includes a mask functionality that allows for precise edits within selected areas and a Magic Prompt feature that automatically enhances and optimizes user prompts for better image quality.

Recraft V3
Recraft V3 is focused on style versatility, letting users choose from a range of preset aesthetics such as Hand Drawn, Pixel Art, and Handmade 3D. It’s built for creators who want to explore distinct visual styles with high-resolution output.

Seedream 3
Seedream 3 supports a wide resolution range, from 512px to 2048px, and offers adjustable quality settings through Small, Regular, and Big modes. It’s also highly reliable for generating text in both Chinese and English, making it a strong choice for multilingual and typographic content.

Luma Photon
Luma Photon prioritizes image fidelity and stylistic accuracy. It supports three types of reference images—Character, Style, and Composition—helping guide the generation process. A Flash version is also available for faster, cost-efficient outputs, with slightly reduced quality.

Plus Additional Improvements & Bug Fixes 🚀

Asset Gallery

• Added a "Create" button with redirection functionality to improve navigation.

Homepage

• Sub-menu Structure: Implemented a new sub-menu layout centered around the "Create" feature.

• Card CTA: Enabled call-to-action buttons on cards for improved user interaction.

Bug Fixes

• Filters: Resolved display issues with the filters header during generation.

• “Explore“ Content: Unified Explore page content for both authenticated and unauthenticated users.

Search

• Thumbnails: Added support for 3D and video thumbnails in search results.

• First Result Display: Fixed an issue that affected the visibility of the first search result.

• Rodin Models: Works with both Image-to-3D and Text-to-3D (using just a prompt).

• Tripo Models: Works with Image-to-3D only.

  

Tripo Family

• Tripo 2.5 supports only one input image.

• Tripo 2.5 Multiview supports four input images from different angles (front, left, right, back).

• Users can set a face limit for meshes.

• Texture Options include PBR material, and Standard vs HD (Max Resolution: 2048x2048).

  

Rodin Models

• Supports up to five input images.

• Concat mode allows to combine images of the same subject and works best when the images share the aspect ratio.

• Fuse mode is designed for blending images of the same subject and does not require them to have matching aspect ratios.

• Mesh densities range from extra-low to high.

• Maximum texture resolution: 2048x2048.

• Generates an optimized geometry for rigging and animation.

• Materials can be PBR or shaded.

  

Seedance 1.0 Lite Update

Now you can use both first and last frames when generating videos with Seedance 1.0 Lite! Try it now on our Video Generation tool.

The upgraded viewer now supports PBR materials, wireframe display, customizable lighting or color settings. Zoom, pan, and rotate with precision. Compare and preview your models as needed before exporting them to Blender, Unreal/Unity, or your 3D software of choice.

Asset Filters

We added filters on all "Asset Gallery“ (your content)  and on the "Explore“ tabs (public content). You can quickly retrieve any asset (whether 2D, video, or 3D) by applying filters like as type, author, date, model, collection, or tags. This makes it easy to find previous work in seconds - even months later.

Explore

You can now use the “Explore” feature as part of the creative flow when generating videos and 3D models. Use the videos here to gather inspiration. You can drag them to your prompt if you are looking to generate something similar!

Navigation & UX

The thumbnail on the Derived Images page now links to the parent asset.

Miscellaneous

• Fixed a glitch showing broken images on 3D and video search results.

• Added a retry mechanism to improve consistency when searching for similar images or videos.

• The "Generate“ button now shows a helpful tooltip when disabled so you can follow the guideline and generate your assets promptly.

• Improved the empty state experience when loading the custom models form.

• Fixed an issue causing an error when retouching assets with prompts.

• Corrected display of parameters for prompt-based edits.

• Improved prompt box responsiveness for smoother editing.

• Fixed a broken link from the Plans page to the Cost Calculator.

You can access our new 3D viewer on our 3D Generation tool. Play with shadows, contrast and get the view you desire. Test it now!

Scenario now supports four new high-performance video generation models:

• Kling 2.1

• Kling 2.1 Master

• Seedance 1.0 Lite

• Seedance 1.0 Pro

These models are now live and ready to use in the Video tool across text-to-video, image-to-video, and hybrid workflows.

🚀 Kling 2.1 + Kling 2.1 Master

• Stronger, faster, more flexible generation

• Great for both realistic motion and stylized creativity

• Kling 2.1 Master gives you full control over aspect ratios:

  • 16:9, 9:16, and 1:1

🌊 Seedance 1.0 Lite & Pro

• Lite version outputs in 720p, optimized for speed and iteration

• Pro version outputs in full 1080p, ideal for production-quality clips

• Both support a wide range of aspect ratios:

  • 16:9, 4:3, 1:1, 3:4, 9:16, 21:9, and 9:21

You can access all four models directly from the Video generation tool, compare outputs across styles, and blend them with other Scenario tools like Expand, Retouch, or Edit with Prompts.

🧪 Try them now and explore the new creative possibilities!

Whether the image is generated in Scenario or uploaded from your device, simply select “Convert to 3D” from the three-dot menu on any image page. You can launch the Image-to-3D tool from the main menu (or access it via direct link.) A dedicated homepage card has also been added for quicker access.

Each image is transformed into a fully-fledged 3D object, complete with mesh and textures. You can tweak parameters like face count, generation steps, and image guidance to control the quality of the result.

Scenario currently supports a range of models, like Hunyuan 2.1 to Direct3D and Trellis. Some models, such as Trellis or Hunyuan MV, also support multi-image inputs for improved accuracy across different angles. More models will be added soon.

Once generated, you can inspect your model in 3D and compare versions created using different generative 3D models. You can then download the result as a GLB file, individually or in batches, with support for additional formats coming soon.

For best results, you may want to remove the background and/or upscale your image beforehand. Clean, stylized inputs usually produce the best geometry and textures.

Image-to-3D is fully integrated with the rest of Scenario, so you can go from a generated concept to a usable asset in just a few clicks.

🔧 Additional improvements and updates:

• Default values for the 3D viewer are now set to brightness 2 and contrast 3.

• The values for the 3D viewer settings are now saved locally, so they don’t reset between visits.

• Paste image functionality has been fixed for both video and 3D views.

• Download and delete have been added to the assets context menu.

• Featured models are back on top of the list.

• Reload prompts have been added to videos.

• The design of the Profile and Organization buttons in the top bar has been updated.

• The Models tab has been moved to the end of the tab list.

We’ve updated the Scenario homepage to improve navigation and usability.

1. The main menu has been moved from the left sidebar to the top of the page, with four core sections.

   • “Create” shows all available tools, offering quick access to each.

   • “Assets“ grants access to your content and custom-trained AI models.

   • “Explore“ allows you to browse the gallery or public assets & models.

   • “Learn“ provides a direct link to the Scenario Knowledge Base.

2. Below the navigation, “Cards“ offers quick access to key feature and highlight updates, news, and case studies as they become available.

3. The “Tools” section has been reorganized by workflow into three main stages (Train, Generate, Edit), with all related tools grouped under each for easier access and clarity.

4. Further down, the Explore section has been enhanced to feature curated creations, reusable assets, and ready-to-use models across different media types & styles.

Overall, the new layout simplifies the experience and better supports the way creators move through their workflows.

We’ve also made updates to how you navigate your account and settings.

• Your “Profile“ menu in the top-right now brings everything together in one place, including Contact, Chat Support, Knowledge Center, Feedback & Suggestions, Community Hub, Trust Center, and more.

• Collections are now accessible via the folder icon next to your profile, making it easier to organize and browse your assets.

• Organization and Project settings remain in the top-left

🔍 New model: Topaz Video Upscale

This release also brings built-in Video Upscale to Scenario. You can now sharpen any video, AI-generated or uploaded, directly inside the platform, with crystal-clear 4K resolution.

More updates to the interface are on the way, with continued UX improvements.

Scenario now includes Flux Kontext, one of the fastest and most advanced image editing models available today. It joins GPT-Image and Gemini as a new model option in Edit with Prompts.

Just upload one or more images, describe or sketch the changes you want, and Kontext will take care of the rest, in seconds.

It’s built for precision and speed, perfect for creating consistent variations or making focused, targeted edits with ease.

You can launch Edit with Prompts from any image ("Edit with Prompts"), from the main menu, or via direct link.

Key Features

• Generate multiple variations for any character, prop, or scene

• Apply targeted changes, from outfit swaps to scene composition

• Support for multiple reference images (up to 3 extra images with Kontext)

• Built-in tools for sketching, cropping, and background removal

• Prompt helpers: generate suggestions, rewrite your input, or translate prompts automatically

• Fully integrated with Video generation, Enhance, Retouch, Expand and other Scenario tools

Visit our Knowledge Base for an introduction, model tips, editing workflows, and more.

Scenario now supports AI-powered video generation (including Google’s Veo 3) with both text-to-video and image-to-video workflows.

You can animate any image, generate a clip from a prompt, or combine both for more creative control. The tool includes over 20 video models covering a wide range of styles, motions, and durations. It’s built for speed and productivity, with an intuitive interface for testing and comparing models, smart prompt assistance, and easy reuse of frames and settings.

Access the Video tool from any image ("Convert to Video"), from the main menu, or via direct link.

Key Features

• Start & end frame animation for controlled transformations

• Looping mode using the same frame for both start and end

• Prompt Spark to rewrite and improve your text prompts

• Flexible settings for resolution (720p / 1080p) and clip duration

• Multiple motion styles: cinematic, stylized, product-focused

• Seamless integration with the other Scenario tools like Retouch, Edit with Prompt, LoRA models & more.

Visit our Knowledge Base for an introduction, detailed model comparisons, best practices, and more.

• We’ve added the latest “GPT-Image” model from OpenAI. This model is among the top-performing models for prompt accuracy and spatial reasoning. It uses uses a particular, unified transformer architecture that natively processes both text and images within a single neural network.

• You can access them via the Base "Models” category on the Model Page (https://app.scenario.com/models) and via our API (https://docs.scenario.com/)

• We’re making three versions of GPT Image available on Scenario:

  • GPT Image 1: OpenAI’s flagship GPT-powered image model, delivering top-tier visual fidelity and precision using the “High Quality” mode.

  • GPT Image 1 (Medium): A balanced mode designed for faster image generation and lower compute cost, with slightly reduced visual detail - ideal for everyday creative workflows.

  • GPT Image 1 (Low): Optimized for speed and iteration, this mode produces visuals or concepts quickly, with minimal resource usage - perfect for rapid exploration rather than final output.

We're thrilled to introduce a powerful new way to create and modify your visuals: Edit with Prompts!

This new tool leverages cutting-edge capabilities of both GPT-Image (aka GPT-4o) and Gemini 2.0, allowing you to make advanced image edits using just simple instructions. You can effortlessly change details, combine elements, alter the mood of a scene, generate consistent variants and more – all with just a few words. Give it a try: https://app.scenario.com/edit-with-prompts

Just upload a reference image and provide related instructions (for example, “create a six different hairstyle variations for this character" or "change the color of the car to blue"). The AI will process your input and generate revised version(s) of your image.

The models are capable of seamlessly performing a very wide range of transformations, such as:

• Add & Remove Elements: Introduce new subjects or remove existing ones from the scene.

• Modify Elements: Change color, texture, size, or style of objects already present.

• Change Backgrounds: Adjust or replace backgrounds, for example, alter the time of day, weather, or overall mood.

• Style Transfer: Apply different artistic styles, like turning a photo into an oil painting or pencil sketch.

• Combine Elements: Upload up to five images to merge elements across them.

• And more!

Beneath the Instructions box, you'll also find helpful options to auto-generate relevant instructions, improve your input with a rewrite or instantly translate prompts into English.

• Invite Users Directly to Projects: Add new organization members directly to specific projects, making the onboarding process faster.

• "Use Image" : Access the "Use image" action across all relevant tools, making it easier to modify images throughout your workflow.

• “Expand“ Jobs in Dashboard: You can now track the “Expand“ (outpainting) jobs in the Sashboard

🛠️ Bug Fixes & Improvements

• Collection Uploads - We've fixed an issue with uploading images directly into collections

• “Generate“ Button is more reliable - Resolved an issue where the Generate button could become disabled after deleting a reference image.

• Gallery Scrolling - Fixed the bug that created unexpected empty spaces when scrolling through a model's generations in “Masonry“ view.

• Canvas Shortcuts - The “ESC“ shortcut now properly deletes masks again, as expected.

• “Retouch“ Tool Enhancement - Re-added missing “Elements” to the Scenario Retouch (Canvas) and introduced color customization options via the context menu.

Introducing “Explore”
Explore is a new way to browse and reuse previously generated content in Scenario - whether it's your own, created by others in your workspace, or public content shared across Scenario.

You’ll find it right next to the “Session“ tab, to the top. Switch to “Explore“, and it will surface content that's relevant to what you’re working on (based on your text in the prompt box), helping you discover new assets, styles, prompts, and more.

From there, you can easily reuse content by copying it or dragging and dropping it directly into your session—then hop back to the Session tab whenever you're ready. With Explore, you can:

• Browse a feed of public generations

• Discover new prompts, styles, and models

• Remix any asset in just a few clicks

Explore is built to speed up your workflow, encourage knowledge sharing within your workspace, spark new ideas, and help you get more out of Scenario.

• Introducing “Audit Trail“ in Dashboards
  Audit Trail is now available for any workspace Admins, inside the Dashboard: offering a clear, centralized view of how you (or your team) are using Scenario.

  Designed with teams and enterprise workflows in mind, Audit Trail helps you understand what’s being created, who’s generating it, and how credits and compute resources are being used across the workspace.

  To access it, head to your Dashboard tab (under “Manage Organization/Project) or click https://app.scenario.com/team?tab=dashboard. From there, you can:

  • Filter activity by date range, user, or job type

  • View a detailed, chronological log of all workspace actions

  • Monitor credit usage and compute costs in real time

  • Identify power users and trends

  Audit Trail currently supports all jobs (training, all image generations and editing jobs plus textures and skyboxes). 3D and video will soon be added to the same dashboard..

  Whether you’re managing small creative teams or overseeing enterprise-scale use as an IT admi, Audit Trail gives you the tools to stay informed, accountable, and in control.

• Improvements & Fixes

  User Experience
  • Added a “Reader“ role to projects (view-only access)
  • Added canvas right-click shortcuts for quicker actions
  • Users now stay in their current view when switching teams (organisations) or projects
  • Team invitees are now prompted to create an account automatically
  • Updated the "Upscale – Balanced" preset for better results

  Bug Fixes
  • Fixed the double-click issue when selecting gallery images
  • Resolved a bug that blocked deletion of pending team invites
  • Non-admin users can now view the team list and leave teams as expected
  • Fixed the issue preventing team deletion on free plans with API keys
  • Fixed a bug where the first canvas generation would appear white

See the difference below and try it out yourself ⚡

• How to Get Started:

  • Go to the "Project" section in the top-left.

  • Click "Create Project" to set up a new workspace.

  • Configure access for your team members.

  • Start creating and organizing content within the project.

• New models gallery on https://app.scenario.com/models/:

  

  • Enhanced filtering and sorting system for quicker model access.

  • The gallery now updates automatically after a model is trained or created – no need to refresh the page.

• New ControlNet “Sketch“ Mode: Added the Sketch ControlNet modality to Flux, designed specifically for hand-drawn sketches or line art references. Use it to transform your sketches into rendered images with your own style models in a streamlined "Sketch-to-Render" workflow.

  

• Controlend Parameter: Introduced the controlend parameter for Flux. You can now decide how long the control will apply to your generation. Start with the recommended settings, and iterate higher/lower as needed to find the right balance.

  

• Model Style Description: Added a feature on the model page that analyzes your training images and creates a style description, in the prompt embedding section. This description is automatically applied to all your prompts, helping maintain style consistency without manual prompt engineering. Especially valuable for difficult-to-train styles or for balancing prompts that might pull toward different aesthetics.

  

• Unified Image Selection Modal: Replaced all image selection modals with the new search-based interface for a more streamlined experience.

⚡ Improvements:

• Canvas Live Performance: faster live inference for Flux squared images.

• Filter Sorting: Authors, models, and tags in filters are now sorted alphabetically for easier navigation.

🛠️ Bug Fixes:

• ControlNet + Img2Img Ratio Restriction: Prevented generation in ControlNet + Img2Img when image aspect ratios differ by more than 20% to ensure output quality.

• Mode Map Transmission: Resolved an issue where the mode map was sent during ControlNet generation even if the toggle was off.

• Grid View Fit: Fixed the 'fit in grid view' functionality to display content correctly.

• Search Autocomplete: Fixed a bug where search autocomplete suggestions appeared even when the search bar was empty.

• Canvas Mask Copy: Fixed a bug where drawing a mask to copy an element in the canvas was not functioning correctly.

AI Boost + Image Search Together – Combine AI Boost with Image Search for even more precise, contextually relevant results.

More Filters Added – Refine search results with an expanded set of filters, offering more control over your asset discovery.

🖼️ Model Page Enhancements

Author Name Display – Model pages now show the author’s name for better attribution and collaboration.

Caption Words Removed – Caption words have been removed from model pages to streamline the interface.

🔍 New Empty State Design for New Users

New users will now see a refreshed, intuitive empty state design that makes it easier to get started and navigate Scenario from the get-go.

🛠️ Bug Fixes & Performance Improvements

✅ Texture Upscale Fix – The upscale function on texture pages now correctly applies texture upscaling instead of image upscaling.

✅ Image Reset Shortcut Fixed – Pressing the "R" shortcut on the Image Details will now reset the zoom as intended, without randomizing the prompt in background when you come from the generation screen

✅ Vectors Display Fixed - SVG format is now the default way to display, rather than JPEG

💡 Got Feedback? We’re Listening!

Your feedback helps us build a better Scenario! Share your thoughts by joining our Discord community, following us on LinkedIn and X, or reaching out directly.

🔗 Join Our Community | Follow Us on X | LinkedIn

Thanks for being part of Scenario—happy creating! 🚀

— The Scenario Team

Our Search feature has been rebuilt from the ground up to provide a faster and more intuitive experience. Whether you're managing a vast library of images or models, locating your desired assets is now a click away.

🚀 How to Access Search

• Navigation Bar – Click "Search" in the top bar.

• Keyboard Shortcut – Press Cmd + F (Mac) / Ctrl + F (Windows & Linux).

📂 Search by Text or Image

• Text Search – Type in keywords to quickly locate relevant assets.

• Image Search – Drag & drop an image or upload one to find visually similar results.

🔎 How Search Works

When searching, our system matches results based on:

• Image Search Fields – prompt, tags, negative prompt, captions, seed, and name.

• Model Search Fields – model name, tags, and caption words.

🎯 Advanced Search Tools

• Filters – Refine results by Type, Author, and Date for precise searches.

• Boost Mode – AI-enhanced results that recognize contextual similarities in images.

• Search History – Quickly revisit and rerun past searches.

Now, finding the right asset is faster, smarter, and more intuitive than ever! 🚀 Learn More

🏗️ Workspace Page Redesign – A Fresh, Modern Look

We've given the Workspace Page a modern makeover! The updated design is cleaner and more intuitive, enhancing navigation and setting the foundation for upcoming features like multi-project management.

🎥 Enhanced Tutorials

We’ve added new tutorial videos covering Pixelate, Texture Generation, and Expand functionalities. Find them in the top bar of your generation page for quick guidance!

✔️ Landscape Edit Reference Upgrade

The mini canvas now dynamically increases in size when in landscape mode, giving you more control when editing.

🖼️ Easier Image & Asset Management

📌 Direct Image Pinning

Love an image? Pin it directly from the zoom view without extra steps! Keep your favorite assets easily accessible.

📊 Period Filter for Better Insights

A new Period Filter has been added to the Dashboard, making it easier to analyze trends and track your creative progress over time.

🖌️ Enhanced Canvas Experience – More Intuitive Controls

✔️ Tooltips on Canvas Elements – Get instant explanations when selecting canvas tools, making navigation smoother.

💡 Got Feedback? We’re Listening!

Your input is invaluable! Join our Discord, follow us on LinkedIn, X, or send us your thoughts directly here. We’d love to hear how these updates improve your workflow!

🔗 Join Our Community | Follow Us on X | LinkedIn

Thanks for being part of Scenario—happy creating! 🚀

- The Scenario Team

Adaptive Fill: Effortlessly remove/erase unwanted elements from your image, directly within your canvas project. More info on our knowledge center article.

• "Vary" Button: Introducing the new 'Vary' button! With just one click, effortlessly explore variations of your creations. The "Vary" feature uses Image-to-Image with a 25 influence, applying the same prompt and generating two images per instance.

  

• “Clear Session” Button: We’ve listened to your feedback and brought back the “Clear Session” button! Say goodbye to clutter and enjoy a clean, organized generation session

• Shortcut List: To accelerate workflows, we’ve added a convenient shortcut list. Master keyboard shortcuts and become a generation pro! Access the shortcut list via the keyboard (press H) or through the top bar menu.

  

• Improved Navigation in the Gallery: In multi-user workspaces, you can now easily find anyone’s work with the new 'Author' filter option, conveniently located at the top of the Gallery.

  

• Skyboxes Now Support Image-to-Image: We've added img2img as a new reference mode for generating 360° skyboxes.

• Custom Textures: Enable a model to generate textures by associating the “Texture” tag on the model page. After tagging it, simply click on 'Use this model/Generate' to directly open a texture generation session.

  

• Download Links in Tasks: We've added download links to the task list, making it easier to export your selected generation without needing to check your email.

• WebSocket-Powered Generations: Image generation is now significantly faster thanks to our transition to WebSockets.

  Various Bug Fixes: We've resolved several pesky bugs to provide a smoother and more enjoyable experience. Thank you for your valuable bug reports!

• “Row View”: Discover a new way to browse your generated content, now with more intuitive filtering! Results are grouped by inference type, and you can refine the view to display only specific Types or Authors (in shared workspaces).

  

• New “Edit“ & “Reuse“ Shortcuts: Use the 3-dot menu on each image to quickly access tools for editing images or reusing their parameters, making the workflow smoother.

  

• Prompt Spark for Skyboxes: Skybox creation just got smarter with tailored prompts designed specifically for 360 skyboxes. Additionally, undo/redo shortcuts in the prompt box make adjustments easier.

  

🐛 Bug Fixes

• Fixed an issue where WebP images could not be sketched on.

• Resolved a bug preventing asset tagging when a tag was already in use.

• Addressed a skybox parameter reload issue where seed values and image counts weren’t set correctly.

💡 Flux Models Update

• Prompt Requirements: Text-to-Image in Flux Pro, Flux Pro 1.1, and Flux Pro Ultra now requires a prompt to ensure higher generation quality.

• New Prompt Box: We improved our Prompt Spark feature

  • Generate a new prompt with the "dice" icon below the prompt box. The new prompt will be tailored to your model, inspired from the captions and any pinned images.

  • Rewrite/expand: Begin with a few words describing the subject or scene. Use the "rewrite prompt" feature to refine, expand, or shuffle suggestions for improvement.

  • Translate: Instantly translate your prompt from another language to English.

  • "Image import“: Drag and drop or upload an image into the prompt box, and its visual elements will be converted into a descriptive prompt.

    

• Style Reference & Character Reference Support for Flux:

  • Style Reference and Character Reference (aka IP Adapter) are now available for Flux Dev, Schnell, and Pro Ultra.

  • You can also combine ControlNet with Style or Character References for Flux Dev and Schnell.

    

• Juggernaut-based LoRAs are now supported. Simply select your base model in the Settings (left panel of the generation screen).

  

• Canvas Enhancements:

  • The mask is now visible in reference images.

  • A new “resize“ option has been reintroduced to the canvas.

• Flux Dev Updates:

  • The maximum number of images per Flux Dev inferences has been raised from 4 to 8.

• Flux 1.1 [Pro Ultra] (the new foundation model from Black Forest Labs), is now available on Scenario, joining the other models of the Flux family. Access [Pro Ultra] in the “Base Models” section or via this link

• Skybox V2: We're thrilled to introduce Scenario Skybox V2, a Flux-based enhanced version our skybox generation model, with improved overall quality, a wider variety of styles, better prompt alignment, and more.

• Tutorial Videos for Restyle and Remove Background: Learn how to efficiently restyle your visuals or remove backgrounds with the newly added tutorials.

• Images can now be downloaded in full quality, regardless of your workspace plan. Batch downloading is also now available to all users.

• Data Deletion: Per our T&Cs, generated content & models will now be automatically deleted one month after a workspace transitions to the free plan. A notification banner will be displayed on assets and in the top bar as a reminder.

Bug Fixes

• Canvas:

  • Fixed an issue where a generation in a small masked area would sometimes be resized to fit the full project dimensions.

  • Resolved an issue where projects weren't fully unlocked after being exited.

• Color Picker: Fixed a bug where the color picker opacity was being set to 0 after using the eye dropper tool.

• Chat with our AI bot: Get instant assistance by asking questions to our AI chatbot! It's an AI assistant trained on documentation, help articles, and other content. Simply use the “Ask AI” feature located at the bottom-right.

  

• Embedded Video Tutorials: Access video tutorials directly on Scenario to learn about various features and tools.

  

• Scribble ControlNet for SDXL: Explore a new way to guide outputs using ControlNet Scribble. Just upload a reference image and select ControlNet Scribble from the dropdown menu (for SDXL-base models only)

• New shortcuts available for Canvas:

  • Quickly change your brush color with the 'c' key!

  • Use the 'shift' key to adjust your brush width.

• Smoother Upload Experience on Canvas: The upload modal now closes automatically after a successful upload.

• Flux Dev on Canvas: Use Flux Dev directly on the Canvas.

• “Use as Reference”: LoRA weights are now maintained when using an image as a reference.

Bug fixes:

• “Mask“ Tool: We've fixed an issue where the mask tool could cause the app to crash.

• Reference Image Updates: The reference image on the canvas is now always up-to-date.

• Canvas Search: Fixed a bug preventing users from opening canvases from search results.

• Collection Visibility: You can now view all your collections, even if you have over 50.

• Enhance: The Precise preset has been updated to produce results that more closely resemble the input image.

• Help modals have been updated

Bug fixes

• Fixed a bug where if you could not see more than 50 collections

• Fixed a bug on the canvas that cause layer to be deleted when you used the eraser hid the layer

• Fixed a bug where the prompt box would become oversized after switching from “Live” to the normal canvas

• LoRA Components in the Canvas: we've reintroduced the "Model Components" sliders in the Scenario canvas, now located at the top-left of the interface under "Model." These sliders are useful for retouching images when working with multi-LoRAs!

• Scenario Live Improvement for Flux LoRA: The “sampling steps” slider has been removed from Flux Live making it now easier to obtain good results with less configuration.

• Collections: You can now create a collection directly from the “Add to Collection“ feature.

• Flux LoRA Training & Multi-LoRAs: Introducing support for training LoRA models on Flux, as well as multi-LoRAs (compositions).

• Flux Live: Added real-time updates and dynamic interactions with the Flux Live feature

• Export Models: You can now export your trained models. Check our documentation for more information on this article.

• ControlNet Integration for Flux Models: Integrated ControlNet for Flux models (excluding Flux Pro). Check our documentation for more information on the new ControlNet modes.

• Reworked Prompt Box: Refined the prompt box by removing tags and introducing an auto-sizing feature.

• Fixed Opacity Issue on the “Mini Canvas“ (Sketching a reference image): Resolved an issue where opacity settings were not correctly applied.

Misc:

• Added “Flux” as a new filter in the model list

• Dynamic thumbnail for skyboxes assets

• Fixed a bug where the lora components would not load after a "use as reference"

• Background Tasks: A new menu in the top bar allows you to monitor and manage both active and completed generations & trainings.

  

• Upscale Training Images: You can now directly upscale (enhance) low-resolution images before starting the model training process.

  

Scenario Live Improvements

• State Persistence: canvas settings — such as influence, prompts, sampling steps, guidance, and seed — will now automatically save when you leave and reload when you return.

• Shortcuts: New shortcuts have been added to the Scenario Canvas & Scenario Live, enabling faster navigation and boosting workflow efficiency.

Bug Fixes

• img2img + Controlnet: Fixed an issue where reloading parameters didn’t correctly retrieve the influences in img2img + Controlnet.

• Thumbnail Uploads: Resolved a bug that was preventing model thumbnail uploads.

• Faster Access: A new tool card and a new link in the left menu have been added for easier access to Scenario Live.

  

• Prompt Spark Integration: The "Prompt Spark" feature is now integrated into Live, offering a powerful and fast method to generate or refine relevant prompts tailored to the custom model being used. More info on our Knowledge Center.

  

• Inference Status Visibility: You can now see the status of your inferences directly in Scenario Live, allowing you to better adjust your input sketching pace (“Queued”, “Initializing Model”, “Generating”

  

• Other Improvements

  • Improved overall performance

  • The canvas drawing opacity is now set to 80% to enhance sketch detail and improve AI output

  • Images exported from the canvas are now labeled as "Canvas Export" for easier identification in your Images Gallery.

  • When clicking "clear all" in Live, the pen tool will stay selected.

  • The outdated "creation model" has been removed: Projects are now created at 1024×1024px with a white background by default, streamlining the process.

Additional Plan Options:

• Credits Auto-Refill for Team and Custom Workspaces: We’ve added a new auto-refill feature for credits in both “Team“ and “Custom“ Plans to keep your resources topped up automatically. More info on our Knowledge Center.

• Add Seats to a Workspace: You can now seamlessly add additional seats to your workspace (“Team“ plan). More info on our Knowledge Center.

  

Introducing Skyboxes (Beta): You can now generate stunning 360° skyboxes directly within Scenario! Whether you're crafting breathtaking panoramic views or immersive environments, creating skyboxes is now simple and intuitive on Scenario.

• Predefined & Custom Styles: Choose from a curated selection of predefined styles or apply your own custom styles to unleash your creativity. For help, check out this Knowledge Center article.

• Create a draft, then Enhance: Start by generating a draft skybox. Once you've found the one that matches your vision, enhance it for the perfect finish!

Texture Upscale: You can now upscale your textures up to 16K, whether they were generated in Scenario or imported.

• Billing information: You can now access your billing information through the “Plans“ tab in your Workspace Settings

  

• Low-resolution warning: A warning will now appear when adding low-resolution images to your training dataset, helping to improve the performance of custom trainings.

• More flexible cropping: Non-square images can now be cropped on the generation interface (in the left column)

• Miscellaneous: In Enhance, the default preset has been changed to "Precise" from "Balanced". We fixed various bugs for improved stability and performance.