Runway Unveils Act-Two: Easy Motion Capture with AI

Runway has unveiled its new motion capture model, Act-Two, which allows users to animate any character—from humans to fantastical creatures—with full detail and realism using just a simple video of an actor.

AI tools are rapidly transforming the video industry, and with the introduction of Act-Two, Runway has taken a major step forward in the field of motion capture. Instead of relying on expensive suits and sensors, this technology only requires two inputs: a regular video of a person speaking, moving, and expressing emotions, and a still image or video of the character you want to animate. The Act-Two AI then automatically transfers all the movements, facial expressions, hand gestures, and even speech patterns from the first video to the second character.

Introducing Runway’s Act-Two AI

Traditional motion capture is a highly complex and costly process that requires specialized studios and expensive equipment. Act-Two has the potential to democratize this process, allowing independent filmmakers, small studios, and even casual content creators to animate characters with a quality close to that of high-end productions.

Built on the Gen-4 video model, Runway’s new system offers significant improvements:

  • Act-Two can simultaneously track head, face, body, and detailed hand movements.
  • It can transfer performance styles to a wide range of characters with different artistic styles—from realistic to cartoon—without compromising the quality or fidelity of the original performance.
  • When using a still image as the character reference, Act-Two not only maps body and hand movements from the source video but also adds subtle ambient motions to make the final result look more natural.
  • When using a video reference, the system preserves the environment and camera motion of the original footage.

Act-Two is now available to users with Standard or higher-tier subscriptions on the Runway platform. The cost is 5 credits per second of output, with a minimum video length of 3 seconds. The system can generate videos up to 30 seconds long in various resolutions at 24 frames per second.