Act-One
Runway's model that transfers an actor's facial performance onto a generated character
Act-One is a feature inside Runway’s Gen-3 Alpha video model, released in October 2024, that turns an ordinary video of a person’s face into a driving signal for an AI-generated character. A creator films themselves on a phone or webcam, and Act-One maps their expressions, eye-lines, and timing onto a separate generated character without any motion-capture rig, facial markers, or manual keyframing. It supports multiple camera angles and can retarget a performance onto characters with different proportions from the original actor.
The feature was aimed squarely at replacing traditional performance capture pipelines that require specialized hardware and studio setups. Runway positioned it as a tool for filmmakers and animators who want expressive character acting without a mocap budget, and it drew attention on release for how well it preserved subtle facial nuance compared to earlier AI lip-sync or face-swap tools. Runway later extended the idea with Act-Two, which adds full-body motion transfer on top of the facial performance capture that Act-One introduced.