FlashML-org/FreeVideo: Run MiniMax H3 video generation locally on consumer hardware

github.com5 min read

FreeVideo is an open-source local inference engine designed to run MiniMax H3 video generation models on consumer hardware with as little as 8 GB of VRAM and 16 GB of RAM. Powered by OpenVDN's Video DeltaNet hybrid attention architecture, it operates as a ComfyUI plugin featuring hardware-adaptive execution, weight streaming, and support for multimodal inputs and community LoRAs across Windows, macOS, and Linux. Open article →

More on this topic