Remove Video Background
Remove the background from footage with AI video matting (BiRefNet v2), leaving the subject on transparency for the whole clip. Spends AI tokens, scaled by input duration.
Category: AI Menu path: AI > Remove Video Background
Ports
| Port | Type | Direction | Description |
|---|---|---|---|
video_in | imageRgba16f | input | Required — connect a Video Source (or another AI video node). The video file is uploaded for processing. |
out | imageRgba16f | output | The matted clip — subject over transparency (delivered as alpha-carrying WebM and decoded back to RGBA). |
Parameters
| Param | Type | Default | Description |
|---|---|---|---|
model | string | (first available) | Single curated model (BiRefNet v2 video, matting variant). |
resultPath | string | "" | Hidden — the saved clip under generated/. |
Standard rows expose Video Source-style playback params (timeMode, speed, reverse, startOffset, loopMode).
How It Works
Async job: the input uploads, the matting model runs remotely (status streams into the button), and the result downloads into generated/. The output composites like any footage — the alpha channel carries the matte, so it drops straight over new backgrounds. Failures refund automatically.
Usage Examples
Background swap
Footage → Video Remove BG → place over a Gradient / generated background layer. Done — no keying, no green screen.
Subject-only effects
Run effects on the matted clip (glow, trails via TrackPoints) while the original footage sits underneath untouched.
Tips
- For a reusable grayscale matte instead of premultiplied transparency, use VideoSegment and wire its matte into track-matte workflows.
- Trim the source clip first — cost scales with input duration.
Related Nodes
- RemoveBackground — single-frame version
- VideoSegment — prompted segmentation when you need a specific object, not the whole foreground
- Matte — combine with other mattes