GitHub radar

ComfyUI H3 clip chaining, super smooth ai video

ComfyUI Plugin for chaining MiniMax H3 AI generated video clips to preserve continuity in motion and audio without a cut or audio drop that would otherwise occur when joining clips.

A ComfyUI custom node pack for MiniMax H3 AI Video Generation Model, which enables video clip chaining. The H3 model can generate both images and sounds simultaneously, however, it doesn’t support video clip chaining. With this plugin, you can pass the last frame’s latent code of previous video clip to the next one, so that the motion speed/direction and sound will be continuous. The patches have been tested and validated at startup, and also compared with the newest version of ComfyUI source code to make sure their validity. When the source code changes and the patches become invalid, they won’t be executed.

Why a vibe-coder should care

With this plugin, you can now easily create AI videos using ComfyUI and MiniMax H3 with smooth and seamless transitions without any jittering or missing sound.

Open on GitHub