Gaurav Misra and Dwight Churchill's Mirage launched Avatar X on Tuesday, an AI video model that the New York company says can build a speaking digital twin from 10 seconds of recorded footage.
https://x.com/trymirage/status/2082120270618530053?s=46 The model powers AI Twin inside Captions, Mirage's video creation and editing app. In a thread on X, Mirage said Avatar X generates a user's voice, facial expressions and body movement together, rather than applying lip movement to an otherwise static avatar. It supports vertical and horizontal video and is designed to maintain a consistent likeness during longer, continuous generations.
Misra and Churchill met while working at Localytics, a mobile analytics company. Misra later led design engineering at Snap, where Mirage credits him with overseeing work on products including Spotlight and Snap Map. Churchill was an early member of Marcus, Goldman Sachs' consumer business. They founded the business in 2021 after concluding that short-form video consumption was growing faster than the tools available to make it.
Avatar X focuses that original thesis on one of AI video's hardest commercial problems: reproducing a recognizable person across scripts, formats and longer clips. Mirage says the model captures small expressions including laughs, yawns, smirks and gasps, with audio driving changes across the face and body. Users can then alter a twin's clothing, background or language without recording another performance.
Mirage's product page compares Avatar X's 10-second enrollment clip with 15 seconds for HeyGen and between one and five minutes for Synthesia. That comparison measures the amount of source footage required, rather than the quality of the resulting avatar. Mirage's broader claims of industry-leading identity preservation and expression are based on its own demonstrations and product comparisons.
The 10-second workflow still matters commercially. A shorter setup process gives Mirage a simpler path to move existing Captions users into avatar generation, where customers can create repeated talking-head videos without returning to a camera. That applies directly to social posts, localized marketing, tutorials and video ads, all formats where producing many variations can matter as much as the quality of a single clip.
From editing app to model company
Avatar X is also the clearest product expression of a transition Misra began formalizing last year. Captions changed its corporate name to Mirage on September 4th, 2025, while retaining Captions as its flagship application. Mirage now describes its business as three connected pieces: the Captions app, proprietary foundation models and an API for other companies generating video in bulk.
Owning the model gives Mirage control over features that would be harder to differentiate if Captions depended entirely on third-party generation systems. Captions supplies an existing distribution channel and editing workflow, while Avatar X gives Mirage a model it can improve around the behavior of creators and marketing teams already using the app.
Mirage has the capital to fund that shift. On March 24th, 2026, it announced $75 million in growth financing from General Catalyst's Customer Value Fund, bringing its reported total funding above $175 million. Mirage said at the time that Captions had more than 20 million users who had created over 250 million videos. Those figures are company-reported, but they explain the distribution advantage behind Avatar X: Mirage can place a proprietary model inside an established consumer product instead of launching it as a standalone research demo.
Avatar X enters a crowded market that includes HeyGen, Synthesia, D-ID and Tavus. Mirage is betting that tighter integration between model, avatar generator and editor will matter more than selling synthetic presenters as a separate tool. The model's performance outside Mirage's selected demonstrations will determine whether that integration produces a durable advantage or another comparable avatar option inside an already busy category.