Hao AI Lab brings FastVideo FastH3 to consumer hardware and releases FastH3 Trim
Hao AI Lab announced that FastVideo FastH3 now runs locally on single consumer machines, supporting NVIDIA RTX GPUs, DGX Spark, and Apple Silicon. According to the lab, FastH3 V2 outperforms base MiniMax H3 in 8 steps, generating a 5-second 480p clip with audio in 15 seconds on a single RTX 5090.
The lab also released FastH3 Trim, a variant 4.2× smaller than base H3 that operates in as little as 8 GB of GPU memory.