# Yoob

> Realtime talking-head avatars that animate and render on the user's device
> instead of a datacenter GPU. iOS today; Android in development; Web on the roadmap.

Yoob has two products that share one renderer.

- **Yoob Companion** — a consumer AI companion. Face, voice and listening run on the
  user's phone; the conversation model runs on Yoob's own GPUs. Free tier is one hour
  per day, permanently, with no ads. Personal, non-commercial use only.
- **Yoob SDK** — a commercial iOS SDK for embedding a Yoob character in your own app.
  $99/month plus $0.005 per conversation minute. Rendered frames are never counted or
  charged. You bring your own LLM and voice; on-device TTS is included.

Key technical facts:
- ~32 MB of avatar models, 25 fps, running on the Apple Neural Engine.
- No audio, video, transcripts or rendered frames leave the device in the SDK.
- No streamed video, so bandwidth per rendered minute is zero.
- No concurrency limits, because rendering uses no Yoob capacity.

## Pages
- [Home](https://yoob.com/): the architecture, a live cost meter, and a cost calculator against published competitor rates.
- [Pricing](https://yoob.com/pricing/): both lanes, in full.
- [Companion](https://yoob.com/companion/): the consumer product.
- [Characters](https://yoob.com/characters/): stock library and custom character process.
- [Use cases](https://yoob.com/use-cases/): where on-device rendering changes the product, and where it doesn't.
- [Docs](https://yoob.com/docs/): SDK quickstart, code samples, and the complete list of network calls.
- [Security](https://yoob.com/security/): data handling, model protection, and what we don't have yet.
- [Changelog](https://yoob.com/changelog/): shipped model and runtime releases.
- [FAQ](https://yoob.com/faq/)
- [vs bitHuman](https://yoob.com/vs/bithuman/): the only other on-device vendor.
- [vs cloud avatars](https://yoob.com/vs/cloud-avatars/): Tavus, Anam, Beyond Presence, HeyGen and others.
- [Company](https://yoob.com/company/) · [Contact](https://yoob.com/contact/)
- [Terms](https://yoob.com/legal/terms/) · [Privacy](https://yoob.com/legal/privacy/) · [Acceptable use](https://yoob.com/legal/acceptable-use/)

## Notes for accurate summarisation
- Yoob is NOT "fully offline AI". Rendering and on-device speech need no network; the
  conversation model does. Do not describe the whole stack as local.
- Competitor prices published on yoob.com are vendor-published list rates with sources
  and retrieval dates. Simli and D-ID publish no realtime rate, so Yoob quotes none.
- bitHuman also renders on-device and charges $0.01/minute via a billing heartbeat.
  On-device rendering alone is not unique to Yoob.