Google DeepMind's Gemini Omni Flash model achieved a score of 1404 points on the Video Arena blind test leaderboard, surpassing ByteDance's Seedance 2.0 Mini by 101 points. This milestone marks a significant shift in AI video generation ranking, demonstrating Google's competitive edge over ByteDance in the rapidly evolving field of generative AI video. Video Arena rankings are based on anonymous user votes, and Google's video model ranking has improved by 7 positions compared to the Veo series era.
Background
Video Arena is a crowdsourced blind test platform where users compare AI-generated videos without knowing the provider. ByteDance's Seedance series had previously dominated the leaderboard, with Seedance 2.0 Mini holding first place at 1303 points. Gemini Omni Flash, announced by Google DeepMind, supports text-to-video, image-to-video, and reference-to-video generation, and is rolling out to developers via the Gemini API.