⬢github Python · 217 ★ +177 since we first saw it · pushed 10 h ago · Apache-2.0
Video-Rebirth/hyperflow-h3-flash
HyperFlow H3 Flash: Video Rebirth's 13.7B, 4- to 8-step compressed MiniMax-H3 for joint video and audio generation (INT8 and GGUF)
HyperFlow H3 Flash is Video Rebirth's compressed, few-step version of the MiniMax-H3 video+audio diffusion model. It generates 5-15 second clips at 1344x768 with synchronized stereo audio in only 4-8 sampling steps, halved model size (~13.7B vs ~31B), distributed as INT8 Python/diffusers weights and GGUF quantizations runnable via ComfyUI, stable-diffusion.cpp, or Unsloth Studio.
Why now: Just announced via a Show HN post highlighting the compression of a 31B video+audio model down to ~14B running in 6 steps, with real-time inference on 4x Blackwell GPUs.
Who it is for: Developers and creators generating video with audio locally, especially those with consumer 24-32 GB RTX GPUs or ComfyUI workflows.
audio-generationdiffusersfew-stepggufint8minimax-h3video-generation
Stars over our 683 snapshots: 40 to 217, since 7 d ago.
Where people talked about it
API: https://socialmediatrends-api.osmike.com/v1/repos/Video-Rebirth/hyperflow-h3-flash