We moved the render to the phone.
Yoob builds realtime talking-head characters that animate and draw on the user’s device instead of a datacenter GPU.
Why this exists
Every realtime avatar product works the same way: a GPU in a datacenter animates a face, encodes it as video, and streams it to your user for as long as they keep talking. It works, and it means somebody is paying by the minute for every second of every conversation. That cost shapes the entire product — it is why free tiers are rationed, why sessions are capped at three minutes, and why engagement is something vendors quietly hope you don’t have too much of.
We spent a long time making the models small enough and the renderer fast enough to do the whole thing on a phone. Face, voice and motion now run on the Neural Engine at 25 frames a second inside about 32 MB. There is no rendering session on our side, so frames are never a unit of account — and there is no concurrency ceiling, because there is no capacity of ours to reserve.
What we believe
- Engagement should not be a liability. If your best users are the ones who cost you the most, your pricing is fighting your product.
- Claims should be checkable. Every number on this site links to where it came from, including our competitors’ prices, and we publish the ones that don’t flatter us.
- Consent is not optional. We do not clone real people’s faces without documented permission, at any price. See Acceptable use.
Where we are
iOS ships today. Android is in development. Web is on the roadmap. Custom characters are a concierge process measured in days. We are early, and the fine print on the home page is a real list, not a humility exercise.
Talk to us
General — hello@yoob.com
Sales and SDK — sales@yoob.com
Security — security@yoob.com
Pricing corrections — pricing@yoob.com