AI & Creative Tools
This Open-Source Video Model Renders in 13 Seconds, But Not Everyone's Allowed to Run It
Hao AI Lab's FastH3 compresses MiniMax H3's video-and-audio generation into 4 steps using distribution matching distillation and sparse attention, hitting a 14x speedup on NVIDIA B200s. Downloads have more than doubled in a day — but the underlying license blocks local deployment in the US, EU, UK, and South Korea.