FastH3 Omni, an open-weight video generation model, has been released, offering significant speed improvements over its predecessor. The model is capable of generating videos using multiple image and video references, along with voice input, to control elements like facial identity and motion. Its training code and reference preprocessing are also publicly available, with benchmarks showing a 5.6x to 11.9x reduction in latency. AI
IMPACT Accelerates open-source video generation capabilities with faster inference and accessible training code.
RANK_REASON Open-weight release of a new generative model with significant performance improvements. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
