MiniMax H3 dropped today with open weights, and ComfyUI has native support on day zero. It’s MiniMax’s third-generation video model (after Hailuo 01 and 02) and the first released open-weights: feed it text, images, video, or audio and it generates video with real stereo sound — up to 2K, up to 15 seconds per clip. Modes include text-to-video, image-to-video, first-and-last-frame control, and reference-to-video, where a reference clip can carry a subject, a motion, or even a voice through the shot. Audio is generated in the same pass, not bolted on afterward.
The engineering story is the local-inference optimization: ComfyUI found the model’s modulation weights (~40% of total parameters) could be pruned and replaced with a functionally equivalent lookup table, added int8 convrot quantization and custom kernels, and cut the total memory footprint 66% — from 123.6 GB full precision to 42.5 GB — so with dynamic VRAM offloading, a 2K video model runs on a GPU like the RTX 3060. Update to ComfyUI 0.30.0 and grab the workflows from the template library; weights are on HuggingFace.