ComfyUI Node
IAMCCS SCAIL Identity Tracker
Draw ordered points/boxes per person on the reference image and driving video, get ref_track_data + driving_track_data out. Play button previews the frames without tracking (partial execution). Auto-detect adds latecomers. IAMCCS variant: use it as a preflight canvas before running the full sampler.
IAMCCS SCAIL Identity Tracker
- sam3_model
- reference_image
- pose_video
- reference_conditioning
- driving_conditioning
- ref_track_data
- driving_track_data
- reference_image
- pose_video
◄refine_iterations2►
◄auto_detecttrue►
◄detection_threshold0.30►
◄detect_interval1►
◄markers{}►
CategoryIAMCCS/video/SCAIL-2
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| sam3_model | MODEL | SAM3 model (e.g. from CheckpointLoaderSimple). | |
| reference_image | IMAGE | Processed reference (post background-removal + padding), at model resolution. | |
| pose_video | IMAGE | Driving/pose video frames, at the resolution fed to the sampler. | |
| refine_iterations | INT | 20–5 | SAM decoder refinement passes per seed. |
| auto_detect | BOOLEAN | true | Master switch for text detection. When on, reference_conditioning / driving_conditioning (if connected) drive SAM3 text detection on that side, alongside any drawn boxes. |
| detection_threshold | FLOAT | 0.300–1 | Score threshold for auto-detected latecomers. |
| detect_interval | INT | 11–64 | Run auto-detection every N frames. |
| markers | STRING | {} | Canvas markers (JSON). Managed by the node's canvas widget. |
| reference_conditioningopt | CONDITIONING | Optional CLIPTextEncode (e.g. 'person') to auto-detect identities on the reference image instead of / alongside drawn boxes. Needs auto_detect on. Text-only loses explicit colour ordering. | |
| driving_conditioningopt | CONDITIONING | Optional CLIPTextEncode to auto-detect / append people on the driving video. Needs auto_detect on. |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| ref_track_data | SAM3_TRACK_DATA | — |
| driving_track_data | SAM3_TRACK_DATA | — |
| reference_image | IMAGE | — |
| pose_video | IMAGE | — |