Updates

This release expands the Video Compositor with transparent backgrounds, audio layers with volume control, audible preview playback, and per-layer matte keying.

Transparent backgrounds

Video compositions can now export with transparency. The background color picker includes an opacity slider — any setting below 100% produces a transparent output. Transparent compositions encode to WebM format with alpha channel support, while opaque compositions continue using the faster H.264/MP4 path. The editor shows a checkerboard behind the viewport when transparency is active.

Audio layers and per-layer volume

Add standalone audio tracks to your compositions alongside video layers. Audio files appear as purple-handled layers with volume controls and timeline trimming. Each layer’s volume can be adjusted individually, and audio mixes into the final render. The editor displays audio waveforms on the timeline for visual reference.

Audible preview playback with sync

Preview playback in the Video Compositor is now audible and stays synchronized between audio and video. The timeline transport includes a master mute control. Preview audio respects each layer’s volume and mute settings, matching the final render output.

Per-layer matte keying

Key individual layers against separate matte sources for advanced compositing. Connect any video source as a luminance or alpha matte to cut out portions of a layer, with an option to invert the matte effect. Matte sources can be reused across multiple layers while maintaining independent timing.

Quality of life

• Click any layer directly on the canvas to select it
• Audio layers show accurate duration with decimal precision
• Preview maintains A/V sync during playback scrubbing

This release fixes a billing discrepancy for reference video jobs and surfaces detailed stage timelines on graph runs.

Accurate billing for reference video jobs

Jobs using reference videos with Seedance 2.5, Seedance 2.0, and MiniMax H3 now correctly bill for the combined duration of reference footage and generated output. Previously, these jobs were only charging for the output duration, which resulted in undercharging compared to the actual provider costs. The pricing now matches the provider’s billing model, ensuring consistent margins across all job types.

The parameter panel shows the updated price estimate when you connect reference videos, so you’ll see the actual cost before running. Jobs without reference videos are unaffected.

Job stage timeline for graph runs

Nodes in graph runs now display the same detailed stage timeline as single-node runs. You’ll see progress steps like “Waiting for GPU capacity” or “Training Gaussian splat” instead of just “Generating…” with an elapsed time. This makes it easier to understand what’s happening during longer-running operations, especially when nodes are waiting for resources.

This release brings higher resolution options to Seedance 2.5 and expands the number of reference inputs you can use for video generation.

Higher resolution options for Seedance 2.5

Seedance 2.5 now supports 1080p and 4K resolutions for text-to-video and image-to-video generation, matching the options already available in the video-edit node. Previously limited to 480p and 720p, you can now generate higher-quality videos directly from text or images. The resolution picker in the node parameters includes all four options, with pricing adjusted accordingly for the higher resolutions.

Increased reference limits for Seedance 2.5

The Seedance 2.5 nodes now accept more reference inputs for video generation. You can use up to 30 reference images (previously 9) and up to 10 reference videos or audio clips (previously 3) when working with text-to-video and video-edit workflows. This allows for more complex multi-reference setups while maintaining the same generation quality and capabilities.

This release introduces new security measures.

This release brings embedded Stripe checkout, clearer media input tooltips, and two new Lucy 3.5 VTON nodes.

Embedded Stripe checkout

Credit top-ups and subscription upgrades now happen inline in a modal dialog instead of redirecting to Stripe’s hosted checkout page. After payment, the modal shows a confirmation spinner while your balance updates, then closes automatically. Your credit balance and billing tab refresh immediately.

Media input tooltips

Media inputs in the parameter panel now show explanatory tooltips when you hover their labels. These describe what each input expects and state any limits—like the maximum number of reference images or supported file formats—so you know the requirements before connecting or uploading.

Lucy 3.5 VTON nodes

Two new Decart nodes are available for virtual try-on workflows: a realtime WebRTC node and an async queued REST API node, both using the Lucy 3.5 VTON model.

Quality of life

• Video Compositor: The layer inspector now shows a preview thumbnail and name of the selected layer, and the layer list highlights the selected row more clearly.
• “Add funds” links now open the billing tab directly instead of the profile tab.

This release improves the reliability of forking nodes with multiple reference images across different providers.

Keep reference images together when forking

When forking a node that uses multiple reference images to a different model provider, all connected images now stay grouped together on the appropriate input handle. Previously, some connections could be silently dropped if the target model used different input naming conventions. This ensures your multi-image workflows maintain consistency when switching between compatible models.

This release strengthens site security.

This release introduces strengthens security enforcement.

This release strengthens platform security to protect user accounts.

This release introduces a new way to compose scenes around your 3D captures and adds a powerful video editing model to the canvas.

Build a scene around your splat

You can now open any Gaussian splat output directly in the 3D Scene editor to create a flythrough video. Use the “Build a scene around it” action on the splat output widget to launch the editor with your splat pre-loaded. The rendered MP4 will appear as an Import node connected to your original splat. This workflow complements the existing splat editor—while that tool focuses on rendering the splat itself at higher resolutions, the scene editor lets you compose your splat within a full 3D environment with primitives and eventually other assets.

Seedance 2.5 Video Edit node

A new Seedance 2.5 Video Edit node is now available. This node lets you edit existing videos using natural language prompts, similar to the Kling O3 workflow. Provide a source video and edit instructions, with optional reference images or audio for guidance. The node supports resolutions from 480p to 4K and includes synchronized audio generation by default. Pricing is based on your source video’s duration, with the final output matching the input clip length.

Quality of life

• The 3D upload limit has been increased from 150MB to 500MB for larger splat and mesh files
• Upload validation now correctly displays the appropriate file size limits for 3D assets
• The CDN now provides progress information when loading large splat files in the scene editor