Updates

Collapsible side panels in the video compositor give you more space for your timeline, and a node lets you generate 3D splats for free.

Meshy 7 3D Generation

A new suite of Meshy 7 models is now available for 3D generation. You can create 3D assets from a single image, multiple images, or directly from text. The node includes parameters for different model types, including standard, low-poly, and smart-topology, giving you greater control over the final output’s style and complexity.

Free Image to 3D Splat

A new node allows you to generate Gaussian splats from images for free. The generated splat is automatically framed in the 3D viewer for immediate preview and can be used in scenes just like any other 3D assets.

This update improves refund handling for failed jobs and adds clearer error messages for content policy violations.

Error classification

Content policy rejections for some models now show as invalid input errors instead of generic failures. This makes it easier to distinguish between policy violations and technical issues. Timeouts during generation are also now properly categorized separately from other failures.

This release adds machine-to-machine authentication support.

This release adds a 3D scene project node to save and revisit scenes on the canvas, a fullscreen editor for the Image Compositor, and a new Seedream model that decomposes images into editable layers.

3D Scene Project Node

You can now save your 3D scene editor sessions as first-class nodes on the canvas. Clicking “Save” creates a project node with a viewport snapshot, and reopening it restores all your edits. The node also supports exporting your scene as video, which can be connected to downstream nodes for further processing. This provides a persistent workflow for iterative 3D scene creation directly within your canvas projects.

Image Compositor Fullscreen Editor

The Image Compositor now features a fullscreen editor, accessible by clicking “Enter Compositor” on the node. This provides a dedicated workspace to move, scale, rotate, crop, and reorder layers using transform and crop gizmos. You can adjust blend modes, opacity, and layer names, with editor-scoped undo/redo. Output settings include canvas dimensions with aspect ratio controls and a background color picker. Export options include PNG, layered PSD, or a ZIP of the raw layers.

Seedream 5.0 Pro Layerize Model

A new Seedream model is available that decomposes an image into a base image and up to 16 transparent PNG layers, each with z-index and bounding box data. The node output includes a widget to recompose the layers at their original positions and export the result as a flattened PNG, a layered PSD, or a ZIP file containing all the individual layers. This provides a structured output for image decomposition workflows.

Quality of life

• The video compositor timeline now uses zoom buttons anchored to the playhead instead of a slider, with a reset button that shows the current zoom level.
• Timeline clips in the video compositor now display thumbnails (first frame for video, the image for stills).
• A new output quality control is available in the video compositor, with presets and a custom bitrate option.
• The video compositor node body is now edge-to-edge, and the “Enter Compositor” button is a full-width footer bar.
• The Jobs panel context menu now includes a “Download assets” submenu for individual artifacts, matching the functionality on canvas nodes.
• The delete button in the compositor layer list is now neutral grey and turns red on hover.

This release adds HDRI support for 3D scenes, makes the video compositor timeline more usable with many layers, and improves output dimension validation. Collapsible panels in the compositor give you more control over your workspace, and previews now show your composition before you start a render.

HDRI support for 3D scenes

HDRIs are now a first-class asset type for use in the 3D scene editor. You can upload .hdr and .exr files directly to the asset library, where they are easily accessible when building scenes. The scene editor’s environment picker is now filtered to show only HDRI assets, ensuring a seamless workflow for lighting your 3D environments.

Collapsible inspector panels in the video compositor

Every section in the video compositor’s right-hand inspector panel is now collapsible. You can hide the Layer, Matte, Transform, Crop, Timing, Audio, and Source sections to focus on the controls you need. Each section defaults to expanded, and the Transform/Crop headers still function as one-click mode switchers for quick adjustments on the canvas. This gives you a more customizable editing workspace.

Pinned timeline header for the video compositor

When your composition has more than five layers, the timeline now keeps the transport bar and time ruler pinned at the top while the layer rows scroll vertically. A scrollbar appears on the right, and the ruler’s horizontal scroll stays synchronized with the clip lanes. This ensures essential controls remain in view, making it much easier to work with complex, multi-layer projects.

Preview frame on the compositor node

The video compositor node now displays a static preview of the first frame of your composition. This preview is generated client-side based on your current layer configuration, giving you a visual check of the output before committing to a render. The preview updates as you edit layers and is replaced by the final video once a render is complete.

Fractional frame rate support

Output frame rates are now selectable from a preset list of standard rates including 23.976, 29.97, and 59.94, with a custom field for entering any fractional value. The presets use exact rational numbers for broadcast-accurate timing, while the familiar rounded values are shown in the interface. This allows for precision when working with specific video standards.

Validation for undecodable video sources

The video compositor now detects and surfaces clips that cannot be decoded by the browser due to unsupported codecs or excessive resolution. An overlay and a warning badge on the layer row indicate the problem, and attempting to render will fail with a clear error message. This prevents sources from being silently dropped from the composition.

This release focuses on developer tooling and internal improvements to lay the groundwork for future user-facing features.

Quality of life

• Minor adjustments to the workspace layout for improved stability.

This release fixes an issue where certain large video and image outputs were incorrectly reported as 2.0 MB in the file properties panel, regardless of their actual size. The true file size is now accurately displayed.

Quality of life

• Correctly display the full file size for large media outputs that were previously truncated to 2.0 MB.

This release fixes a critical issue that prevented older generation jobs from loading or rerunning correctly, ensuring compatibility with recent model updates.

Restored compatibility for older generation jobs

Jobs created with older model versions, such as Kling O3 video generation nodes, will now correctly resolve their type and inputs when viewed in the jobs list, opened in the canvas, or rerun. Previously, these jobs appeared broken with missing icons and incorrectly labeled parameters. The system now automatically migrates the stored job definitions to their current equivalents, preserving your ability to work with past generations.

• Jobs panel no longer shows question mark icons for older Kling O3 and Seedance 2.0 jobs
• Input parameters are correctly mapped when reopening older jobs in the canvas
• Rerunning jobs from the feed now uses the correct model and parameters

This release adds a user-controlled threshold for when node metadata becomes visible on the canvas.

Customizable node metadata visibility

You can now control the zoom level at which node metadata—such as resolution, duration, and model badges—appears on the canvas. A new setting, Show node metadata above zoom, is available in Preferences → Canvas. The value is a percentage from 10% to 800%, defaulting to 80% (the previous fixed behavior). Setting it to 10% keeps the metadata visible at all zoom levels, while 800% effectively hides it.

This release adds support for uploading image sequences as ZIP files to Gaussian Splatting nodes and introduces layer removal shortcuts in the video compositor.

Upload image sequences as ZIPs to Gaussian Splatting nodes

You can now upload a ZIP file containing an image sequence directly to the image input on multi-view Gaussian Splatting nodes. This eliminates the need to extract and manually select hundreds of individual photos when working with captured sequences. The ZIP upload is mutually exclusive with individual image uploads or video inputs—you can use only one method at a time. The system handles the archive upload efficiently and passes it directly to the backend for processing.

Remove layers directly from the video compositor

Layer management in the video compositor is now more direct. An inline remove button (X) has been added to each layer row in the layer panel, and a “Remove layer” option is available in the viewport’s right-click context menu. Additionally, the Delete or Backspace key will now remove the currently selected layer instead of deleting the entire compositor node, making layer cleanup faster.

Quality of life

• Pressing Delete or Backspace with a layer selected in the video compositor now removes that layer instead of the entire node.
• The segmented toggle controls for enum fields now have proper accessibility labels for screen readers.
• Transparent WebM exports are now processed in chunks to prevent out-of-memory errors on longer compositions.