ComfyUI Node
MiniMax H3 Transplant Audio Latent
Transplants the completed audio stream from one MiniMax H3 joint AV latent into another. The target video is left unchanged, video is fully denoised, and audio is fully preserved. Audio shapes must match exactly; the node never crops, pads, resamples, or converts dtype.
MiniMax H3 Transplant Audio Latent
- source_av_latent
- target_av_latent
- LATENT
Categorymodel/latent/minimax
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| source_av_latent | LATENT | Completed MiniMax H3 AV latent whose exact audio stream will be preserved. | |
| target_av_latent | LATENT | Fresh MiniMax H3 AV latent providing the target-resolution video stream and metadata. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| LATENT | LATENT | Target AV latent with source audio transplanted, video mask 1 (denoise), and audio mask 0 (preserve). |