Is MiniMax H3 The New Lip Sync King? MiniMax H3 is live on OpenArt! A cinematic AI video generator that takes text, image, video and audio references in a single generation, with native stereo sound. Full walkthrough and my own tests below.
H3 outputs native 2K at 24fps, 5–15 seconds per clip. What makes it different from most AI video generation tools is the prompt: instead of picking a mode, you describe how your assets relate to the shot you want — camera move from one clip, character from an image, vocal from an audio file, in one sentence.
In this video I cover text to video with native stereo audio, image to video using first and last frame, reference-led generation for realistic AI video, video-to-video motion transfer, and instruction-based AI video editing — changing an object, the lighting, the pacing or the sound without regenerating the whole clip.