Video Arena is now available on the web, allowing users to create videos using 15 AI models and comparing their outputs side-by-side from the same prompt.

Users generate videos and vote on which outputs they prefer. These votes are aggregated into public leaderboards intended to reflect real-world model performance across diverse prompts and use cases.

Check out the rankings Text-to-video Leaderboard: https://lmarena.ai/leaderboard/text-to-video Image-to-video Leaderboard: https://lmarena.ai/leaderboard/image-to-video

Which prompts seem to differentiate models well? Which didn't? Are there any surprising results?