Captions launches Mirage Studio with expressive AI actors

Most AI video avatars still suffer from the same problem: they talk like wooden mannequins reading from a teleprompter.

Video creation startup Captions is trying to fix that. On June 2, the company launched Mirage Studio, a new creative tool powered by its proprietary omni-modal foundation model.

Directing expressive digital actors

The workflow is straightforward. Creators can upload an audio track, describe the intended scene, or supply a reference image to generate energetic videos in minutes.

Captions claims its virtual actors actually feel alive. Instead of just stiff lip-syncing, they can laugh, flinch, sing, and rap based on textual direction. The studio is now publicly accessible at mirage.app.

Why it matters

For marketers, solo creators, and small narrative teams, producing dynamic video without hiring cast or rigging 3D models has always been the holy grail. If Mirage can consistently deliver genuine micro-expressions and rhythmic movements, it removes a massive bottleneck in short-form content and ad creation.

That said, we think the gap between curated demos and everyday output is notoriously wide in expressive AI video. Whether these digital actors can maintain physical consistency across full scenes without uncanny visual glitches remains the real test.