GitHub radar
H3 clip chaining for ComfyUI: seamless AI video
ComfyUI plugin that chains MiniMax H3 AI-generated video clips so motion and audio continue seamlessly across each join, without the seam or audio reset that normally break long takes.
A ComfyUI custom node pack that enables clip chaining for the MiniMax H3 AI video generation model. H3 generates picture and audio together but lacked built-in chaining; this plugin feeds the tail latent of the previous clip directly into the next one, keeping motion speed, direction, and audio continuous across the join. The patches verify their own math against the live ComfyUI code on each startup and refuse to run if an update breaks an assumption.
Why a vibe-coder should care
If you generate AI video in ComfyUI with MiniMax H3, you can now build scenes that flow into each other rather than restarting — longer takes, consistent motion, and audio that actually continues rather than sounding like a cover band starting from scratch each time.
▌ More finds