AI avatar model
HumanLike Homo 1
HumanLike$0.005 a minute, 300ms end to end. The price disruption is the product.
Homo 1 is our own real-time avatar. It does one thing — put a convincing human face on a live conversation — and it does it at a price that changes what you can build. At $0.005/min an always-on avatar costs less per hour than the electricity to display it, which makes it viable for surfaces nobody would put a $1/min avatar on: every job posting, every candidate FAQ, every product page.
Sample output
What it actually produces.
One clip is worth more than any benchmark row. Where we have not published a sample yet, the frame below holds the space it will occupy.
Demo clip
No sample of HumanLike Homo 1 uploaded yet.
What it animates
Head to toe.
A model that only drives the mouth looks wrong the moment the other person starts talking — nothing on screen moves. Filled means driven, outlined means limited or looped, greyed means static.
- Face
- Full
- Lip-sync
- Full
- Upper torso
- Partial
- Hands
- None
- Full body
- None
Face and lips are fully driven at 99% lip-sync. Torso carries a subtle idle motion so the frame never freezes between turns; hands and full body are out of frame.
Source: HumanLike Homo 1 rendering pipeline
Variants
1 way to run it.
Same family, different quality/cost rung. Picking the wrong one is the most common way teams overpay.
Real-time
Balancedhomo-1-realtimeLive conversational rendering over WebRTC at 300ms end-to-end.
Live candidate chat · Always-on site avatars
Fit
Two audiences, one model.
The same model is a different proposition depending on what you point it at.
Recruitment fit
The cost floor for candidate-facing video — cheap enough to leave running on every job page, not just the interview.
Best for
- Job-page avatars
- Candidate FAQ and chat
- High-volume first contact
Strengths
- Two orders of magnitude cheaper per minute than CVI-class providers
- 300ms latency reads as conversational rather than laggy
- Native to the platform: no third-party data processor in the candidate path
Watch out for
- No visual perception layer — it does not see the candidate the way Tavus Raven does
- Proctoring and attention signals need a separate component
The numbers
Measured, not asserted.
Published benchmarks
Vendor and independent figures.
- Lip-sync accuracy
- 99%Our eval
- End-to-end latency
- 300 msOur eval
Compare
