ComfyUI Node
MiniMax H3 Chain De-Rope Budget
Report De-Rope cost before allocating the expanded IMAGE batch. Does not cap frames, change the hold map, or reject an expensive plan. The pixel-buffer estimate excludes endpoint/grid padding, copies, later spatial upscaling, model weights and activations; it is not a VRAM forecast.
MiniMax H3 Chain De-Rope Budget
- state
- source_video_latent
- sigmas
- hold_map
- sampling_steps
- status
◄hold_map—►
Categoryconditioning/minimax/context_loop/upscale
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| state | H3_CHAIN_UPSCALE_STATE,H3_CHAIN_STATE | Current scene state; uses its RAW clock, including the carried prefix. | |
| hold_map | STRING | Protected hold map from Chain De-Rope Guard. Passed through unchanged. | |
| source_video_latent | LATENT | Clean source video latent, used only to report the source canvas. | |
| sigmas | SIGMAS | Actual truncated sampling schedule. Its length determines the real step count. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| hold_map | STRING | Unchanged protected map; connect to H3 Time Smear so this report runs before expansion. |
| sampling_steps | INT | Number of sigma intervals; connect to Time Smear est_steps to keep its estimate current. |
| status | STRING | Planned expansion, actual steps and source-resolution pixel-buffer estimate; connect to Preview Any. |