ComfyUI Node
🐧 SAM3 Video Track (Mask Index)
SAM3 video tracking seeded from a mask on any frame instead of only the first. Tracks forward from start_index and, with bidirectional enabled, tracks the earlier frames in reverse and merges both halves. Output mask count always matches the input frame count; uncovered frames are empty. For text-prompted detection use the native SAM3 Video Track node.
🐧 SAM3 Video Track (Mask Index)
- images
- model
- mask
- track_data
◄start_index0►
◄bidirectionaltrue►
CategorySuperNodes/Video
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | Video frames as a batched image [B,H,W,C]. | |
| model | MODEL | Loaded SAM3 model. | |
| mask | MASK | Mask(s) for the start_index frame to track (one per object). | |
| start_index | INT | 00–99999 | Frame the mask belongs to (0 = first frame, matching the native node). Tracking runs forward from here. |
| bidirectional | BOOLEAN | true | Also track backwards from start_index to frame 0 by reversing the earlier frames. Off leaves those frames empty. Ignored when start_index is 0. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| track_data | SAM3_TRACK_DATA | Track data for SAM3 Track to Mask / SAM3 Track Preview. |