MiniMax H3 — Free Open-Source AI Video Model

YouTube Short (:33)

Claims

• Chinese free/open-source AI video model beating Google’s model on leaderboards; described as uncensored
• Text-to-video, image-to-video, or video-reference generation
• Auto-generates audio in one pass: realistic speech, lip-sync, scene-fitting music
• Runs locally on an average PC with 8 GB VRAM using their optimized variant