NikoDemon80 /ComfyUI-H3-Motion-Context

About ComfyUI-H3-Motion-Context

Chain MiniMax H3 clips into longer sequences with this ComfyUI node pack. It pins the last frames and audio of one clip to the head of the next. Audio keeps going instead of restarting, which is the harder half to get right. Quality drops down a long chain, mostly in the audio.

Best for: Editors assembling multi-clip MiniMax H3 sequences who need motion and sound to carry across each cut.

• Pins the previous clip's tail frames from its latent, so no decode or re-encode
• Continues audio by pinning a window that ends at the join and reaches backwards
• Removes the pinned head with a Trim node, picture and sound together
• Carries a clip between runs with Save Latent and Load Latent nodes
• Loops a chain with Run/Re-roll, Approve, Chain, Reset and Clear latents buttons
• Needs ComfyUI 0.34.0 or newer; version 0.3.1 runs on older builds

1. 1
   Copy the folder into ComfyUI/custom_nodes/ and restart ComfyUI.
2. 2
   Check the startup log says the nodes registered; layout checks run on first use.
3. 3
   Wire the H3 image-to-video, reference-to-video or t2v node into H3 Motion Context, then to the guider/sampler.
4. 4
   Carry the previous clip across runs with Save Latent and Load Latent (first clip: Load 0, Save 1).
5. 5
   Wire trim_frames into H3 Motion Context Trim before Create Video or save.
6. 6
   Put Load, Save and Chain in one canvas group, then use the Chain node's buttons.