AI
MiniMax launches H3 multimodal video model

H3 accepts text, image, video and audio context and generates 4–15-second video at up to 2K with native stereo audio. MiniMax says it supports editing, reference-driven generation and motion transfer, and plans to release weights subject to law and regulation. Official list pricing is US$0.13 per second at 2K and US$0.09 per second at 768p, with 768p in closed beta.
Impact: The release increases price pressure in commercial video generation and gives media, advertising and design teams a more controllable multimodal workflow. Claims about comparative quality and cost remain vendor-reported.




