What Liva AI does
Liva AI is a data company building multimodal datasets of human voice and video to help voice and video AI models feel more human. It collects naturally occurring, rights-cleared human audio and video and delivers it to research teams building next-generation voice models. Liva AI was founded in 2025, is headquartered in San Francisco, and is backed by Y Combinator.
Key capabilities
Liva captures real, consented human voice and video entirely in-house, using no synthetic data or third-party purchases, so all content is authentic and rights-cleared. Through consumer-facing software, crowdsourcing, and partnerships, it gathers diverse accents, emotional range, and varied contexts, including sales calls, multi-channel dialogues, expressive monologues, casual conversations, and interviews, at high production quality.
Who it's for
Liva AI serves research and engineering teams training voice and video AI models that need diverse, high-quality, legally clean training data, addressing the gap left by web scraping and existing datasets that lack coverage across accents, emotions, and languages.