ComfyUI Node
MiniMax H3 Guided Continuation Window
Allocates a fresh AV target whose leading overlap is generated as hidden context under a MiniMax H3 Latent Tail Guide.
MiniMax H3 Guided Continuation Window
- previous_av_latent
- latent
- window_length
- overlap_video_tokens
- overlap_audio_tokens
- transition_video_tokens
- transition_audio_tokens
◄overlap_frames22►
◄extension_frames119►
CategoryMiniMax H3/continuation
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| previous_av_latent | LATENT | The completed cumulative H3 AV latent to continue. | |
| overlap_frames | INT | 225–3600 | Hidden motion-context span; must satisfy 17k + 5. |
| extension_frames | INT | 11917–3570 | New visible frame count; must be a multiple of 17. |
Outputs (6)
| Name | Type | Description |
|---|---|---|
| latent | LATENT | Fresh unmasked AV target for the continuation sampler. |
| window_length | INT | Connect to MiniMax H3 Image to Video's length input. |
| overlap_video_tokens | INT | — |
| overlap_audio_tokens | INT | — |
| transition_video_tokens | INT | Always zero: the complete hidden overlap is discarded after sampling. |
| transition_audio_tokens | INT | Always zero: the complete hidden overlap is discarded after sampling. |