Movies in every language — AI dubbing with identity permanence
HQ: Toronto, ON, Canada | Founded: 2026 | Employees: 8 | Stage: Seed (Y Combinator S26) | Website: https://thefamiliarlab.com
Familiar calls itself the 'Neolab for world translation', dubbing film and video into 30 languages. Its technical claim is about how the model is built: rather than chaining separate systems for lip-sync, speech synthesis and scene handling, one model produces the voice, the lips and the scene together, keeping a consistent identity and an understanding of the 3D scene. In practice that means the same actor's performance is preserved instead of replaced by a mismatched voice on a mismatched face. The company says its output matches human translators on a published test, hosted at thefamiliarlab.com/benchmark. Founded in 2026 and based in Toronto, Canada, with 8 employees. Chief executive An Zhu Liu is a Stanford dropout who sold voice AI for six figures, built what he describes as the best human avatar cloning model of the spring, and turned down an offer to lead model research at a voice company to start Familiar. Mingi Kwon holds an AI PhD from Yonsei University, built large-scale text-to-video models at Adobe Research, and previously ran localised YouTube channels with more than 400,000 subscribers. Xu Zheng holds an AI PhD from HKUST with more than 30 publications, over 2,000 citations and more than 2,000 GitHub stars.
Dubbing is one of the few uses of AI in media where the output is easy to judge and the buyer already has a budget: studios and streaming services spend real money on translation, and the current result — a mismatched voice on a mismatched face — is visibly bad. Familiar claims to solve it in one model rather than a chain of separate lip-sync, voice and mixing steps, so the actor's identity and performance survive translation. The founding team is the strongest signal: a text-to-video researcher from Adobe Research, an HKUST PhD with more than 2,000 citations, and a Stanford dropout who had already sold voice AI commercially. The risk is structural — localisation is a per-project services market and studios buy slowly.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki