Physion Labs has launched Physion-Arc 1.0, a new benchmark designed to evaluate whether video agents can create complete, multi-scene videos—not just visually impressive individual clips.
The benchmark assesses 600 videos generated from 100 screenplays across narrative coherence, cinematic language, visual continuity, pacing, direction, and overall production quality. It includes leading video-generation platforms such as Runway, Luma, MiniMax, Kling AI, Utopai Studios, and TapNow.
Runway Agent 2.0 ranked first overall and led every evaluation dimension, particularly in subjective areas where cinematic judgment and storytelling quality matter most.
Physion-Arc represents an important step toward measuring video models as creative and directorial systems—helping the industry understand whether generated scenes can come together as a coherent and emotionally engaging experience.