Pricing
Usage-based pricing is advertised from as low as $0.06 per minute. Full plan tiers are not listed on the homepage, and sales contact is offered for larger deployments.
Learn morePhotoreal conversational avatars with real-time latency for developers

Keyframe Labs is an avatar tool. Keyframe Labs offers photoreal, real-time conversational avatars driven by its own persona-1.5-live model. Developers pair them with any LLM or voice model through no-code components or a JavaScript SDK, with pricing from $0.06 per minute.
Keyframe Labs builds human foundation models and uses them to power interactive avatars that talk with users in real time. Its first model family includes persona-1.5-live, a 200M parameter model aimed at realistic faces, accurate lip sync, natural motion and a wide range of facial emotion. The result is a video avatar that can sit on top of a voice agent or chatbot and give it a face. The product is aimed at developers and teams that already have a language model or voice stack and want a visual layer for it. Typical uses include customer-facing agents, support, sales and training experiences, and enterprise deployments that need many simultaneous sessions. The company positions the platform for businesses that care about both visual quality and scale, rather than for casual video creation. In practice, setup follows three steps: pick an avatar, connect any LLM or voice model, and go live. The homepage lists compatibility with providers such as OpenAI, ElevenLabs, Cartesia, Gemini, Grok, Hume and Rime, so the avatar acts as a rendering layer rather than a locked-in brain. Teams that want more control can use the JavaScript SDK, which opens an avatar session, receives video and audio tracks for HTML elements, and accepts audio bytes from the team's own model. A LiveKit option is also shown, and docs and example projects are published. An image-to-avatar feature turns a portrait into a usable avatar, and an emotion control system lets builders steer expressions such as happy, skeptical or confused. Performance claims center on a 180ms time-to-first-byte, compared with 450ms or more for other providers according to the vendor's own chart. Enterprise features include thousands of concurrent sessions, multi-region servers in the US, EU and Asia, and a zero data retention mode. Among alternatives, it competes with other real-time avatar and video agent APIs; its distinguishing points are its in-house model research, emotion controls and per-minute pricing starting at $0.06.



Usage-based pricing is advertised from as low as $0.06 per minute. Full plan tiers are not listed on the homepage, and sales contact is offered for larger deployments.
Learn moreDocumentation and example projects on GitHub are provided, and a Talk to Sales option books an intro call for enterprise needs.
Learn moreWorks with LLM and voice providers including OpenAI, Cartesia, Grok, ElevenLabs, Gemini, Fish, Qwen, Sonnet, Hume and Rime. Offers a JavaScript SDK with client, server and LiveKit options.
Learn morePhotoreal avatars with lip sync, natural motion and hair and lighting physics; 180ms time-to-first-byte; broad emotion control; image-to-avatar; support for many LLM and voice providers; JavaScript SDK; unlimited call length; zero data retention mode.





