Node reference
Video tools
Six nodes that work on a clip you already have: cut it, pull a frame from it, describe it, remove its background, grade it to a reference, or make it bigger.
Last reviewed October 25, 2026
Tools for footage
The video tools take one clip on their input and give one result on their output. Two are free and run as lenses with no Run button. One writes words. Three send the clip to a model and spend credits, and those three price by the clip's length, which is why trimming first is the cheapest decision on the canvas.
| Tool | Takes | Gives | Price |
|---|---|---|---|
| Trim | Video | Trimmed Video | Free |
| Extract Frame | Video | Frame Image | Free |
| Video Describer | Video | Prompt | Credits, by clip length |
| Alpha Matte | Video | Output Video | Credits, by model and length |
| Video Color Match | Video, Reference | Output Video | Credits, by model and length |
| Video Upscale | Video | Output Video | Credits, by model and length |
Trim and Extract Frame are free lenses
- Trim
- Set the window on the filmstrip bar and the Trimmed Video output means that segment. Nothing is cut until a node downstream runs; then the editor cuts the clip first, for free, and hands the segment on. Download renders the file on demand at full size.
- Extract Frame
- Pulls one frame out as an image. The playhead scrubs to the exact frame and the step buttons move one measured frame at a time. Last is the default, because the last frame of one clip is the usual start frame of the next.
Trim before anything paid
Video Describer
Video Describer watches a clip and writes one description that would recreate it from words. It follows Image Describer exactly: an action button on the node starts it, the words stream in, the text stays editable, and the Prompt output carries your edited version. It describes only what is visible; there is no creativity setting, because a faithful description and creative licence cannot both be true.
It is priced by the clip's duration, roughly a picture's worth per second. A clip longer than the model's window is refused with the honest fix: trim it first.
Alpha Matte
Alpha Matte removes a clip's background and returns the subject on transparency, ready to composite over a new scene, or the matte itself where the model offers it. It is matting, not generation: the footage is never re-rendered, only given a per-pixel opacity, so faces and motion stay exactly as filmed. Its models live on this node only, and the dials shown are the ones the selected model has.
Video Color Match
Video Color Match grades a clip to a reference picture's colors and lighting through an edit-capable video model. The clip goes on Video, the reference on its own Reference port, and two toggles decide what to match. There is no prompt field: it is written on the server from the toggles. The output keeps the clip's length, shape and every frame's content and motion.
A clip with a recognisable face
Video Upscale
Video Upscale gives a clip more pixels with the same motion. It is the node with no prompt, handed a clip and asked for more of it. Pick the model and the target size it offers; the quote on the Run button is what the hold takes.
Common connections
Video Input → Trim → Video Upscale
Pay to upscale only the seconds you keep.
Video Preview → Extract Frame → Video Generate Start Frame
Continue a shot from the frame it ended on.
Extract Frame → Image tools
Pull a still and treat it like any picture: crop it, describe it, upscale it.
Video Preview → Video Describer → Video Generate prompt
Recreate the motion and mood of a result with a new subject.
Video Input → Alpha Matte → Video Generate
Lift a filmed subject onto transparency and place it in a generated scene.
Image Preview → Video Color Match Reference
Grade a clip to match a campaign's hero still.