Recipe: Edit & clean up video
Goal: take raw footage and transcode it, split it into frames/segments, or remove an unwanted object from the shot.
Nodes you'll use: Video Input → Video Converter / Video Split / FFmpeg GPU Video Process, and for object removal SAM2 Video Mask → Video Object Remover / Video Painter.
Transcode or split
- Add a Video Input node pointing at your clip.
- Add Video Converter (change format/codec) or Video Split (to frames or segments).
- Connect, Save, Run, download.
For heavy jobs, FFmpeg GPU Video Process runs the transcode on a GPU worker.
Remove an object from a shot
- Video Input → a masking node (e.g. SAM2 Video Mask) to select the object across frames.
- Feed the masked video into Video Object Remover (or Video Painter) to inpaint where the object was.
- Run, then download the cleaned clip.
Object removal is frame-by-frame inpainting and is GPU-heavy — expect longer runs on longer clips, and watch the per-node progress.
Over the API
Same loop as any workflow — upload the clip, run your saved video workflow:
# $AUTH is "Authorization: Bearer <token>" — get it from your sk_ key,
# see /api-reference#authentication
COG=$(curl -s -X POST "$BASE/cogs/upload-file" -H "$AUTH" \
-F "file=@./clip.mp4" | jq -r .id)
curl -s -X POST "$BASE/workflows/workflow-versions/<your_version_id>/run" -H "$AUTH" \
-H "Content-Type: application/json" -d "{\"cog_ids\": [$COG]}"
Related
- Node Library → Video
- Batch-process a folder — run over a whole library of clips.