An Open Platform for Evaluating LLMs by Human Preference
VideoArena establishes the first dynamic leaderboard for SOTA text-to-video generation models. Our mission is to study human preferences on AI-generated videos, decoupling prompt quality from model capabilities, and understanding the impact of realistic dynamics vs photorealism. By taking design inspiration from social media platforms like TikTok and YouTube Shorts, VideoArena aims to create a simultaneously engaging and informative interface to evaluate AI generated videos.
Contributors
Michael Luo, Justin Wong, Brandon Trabucco