China’s MiniMax Launches New Open-Source AI Video Model
MiniMax is open-sourcing its flagship video model in a challenge to ByteDance, Google and Kuaishou’s Kling.
Chinese AI firm MiniMax on Friday announced a new open-source video generation model, H3, cranking up competition against AI video rivals like ByteDance and Google.
MiniMax is trying to shake up the market by open-sourcing its flagship video model for the first time. MiniMax said H3, a multimodal model that understands text, image, video and audio in prompts, can generate videos up to 15 seconds long with stereo sound.
Also on Friday, ByteDance made its latest video generation model, Seedance 2.5, widely available to the public through its Dreamina AI image and video platform. ByteDance first unveiled the model last month, but it was in beta testing until Friday’s official release.
MiniMax’s H3 ranks second in the text-to-video category, just behind Google’s Gemini Omni Flash, and first in the video-editing category on a leaderboard compiled by Artificial Analysis. Seedance 2.5 doesn’t appear on the leaderboard yet.
Apart from Google, most major developers of AI video generation models are Chinese, including ByteDance, Kuaishou’s Kling AI video unit, MiniMax and Alibaba.