Runway releases Act-One performance capture

Background

Traditional facial animation pipelines relied on motion capture equipment, multiple footage references and manual face rigging^[2]. In practice that means several distinct stages: a capture setup that records the performer, a body of reference footage to work from, and an animator building and correcting the controls that move a character's face. Every stage sits between the actor's performance and the finished shot, and every stage adds cost and time.

What was released

On 22 October 2024, Runway released Act-One within Gen-3 Alpha^[1]. The tool takes a video of an actor's performance together with the voice, and turns it into an animated, generated character^[1]. The input is a performance recorded on camera; the output is an animated character that carries that performance.

What is different

Runway's claim for Act-One is that it needs no motion-capture equipment and no manual face rigging^[1]. Both were standard parts of the pipeline described above^[2], so removing them changes what a production needs before it can begin: a camera and a performance rather than a capture stage and a rigging pass.

Runway said eye-lines, micro expressions, pacing and delivery were faithfully represented in the generated output^[2]. These are the details audiences read as acting. The eye-line tells you what a character is looking at, and whether they are meeting someone's gaze or avoiding it. Micro expressions are the small, fast changes around the eyes and mouth that register a reaction before a character speaks. Pacing and delivery cover the timing of the performance and the way the line is voiced. Getting these right is what separates a character that appears to perform from one that merely moves its mouth in time with a soundtrack.

Later release

A later tool, Act-Two, became available on all paid plans^[1]. The capability that arrived with Act-One inside Gen-3 Alpha was thus extended to Runway's paid subscription tiers.

Why it matters

Face work has long been the part of character animation most tightly bound to specialised equipment and specialist labour^[2]. A technique that accepts ordinary video as its input moves the starting point of that work closer to what most productions and individual creators already have: a camera, an actor, and a take. For dialogue-driven animation, the performance itself becomes the thing you record, rather than a description of the performance that an animator later rebuilds by hand.

It also changes the order of the work. When the performance is captured as video and translated directly, the acting decision and the animated result are separated by fewer steps, and the faithful reproduction of eye-lines, micro expressions, pacing and delivery^[2] becomes the measure by which the result is judged.