Runway Act Three brings facial mocap to any actor

Runway shipped Act Three this week — the third generation of their actor-driven video tool. You upload a target character (image or generated video), record yourself performing the scene on a webcam, and Runway transfers the full performance: facial expressions, eye movement, head pose, and lip sync.

Why we picked it

Act One was a tech demo. Act Two was the first version that worked. Act Three is the first version where the output crosses the uncanny valley reliably — at least for talking-head shots in good lighting.

Key updates

• Eye-tracking now produces natural saccades, not the "dead doll" stare from v2
• Multi-character scenes — animate two characters from two simultaneous webcam feeds
• Output up to 1080p (was 720p in v2)
• Pro plan gets 90 seconds of unwatermarked render per session

Watch out for

• Heavy lighting and shadow scenes still produce artifacts on the chin and ear edges
• Side profiles below 30° break the lip-sync model — keep cameras roughly forward
• Free trial is 30 seconds total; not enough to evaluate seriously

For solo creators making short-form video, this is a real production tool now, not a novelty.