Compare to other avatar APIs
Pricing, realtime sessions, session length, and architecture across the market.
See comparison →◆ Enterprise
Realtime avatar infrastructure
for enterprise AI products.
Add photorealistic realtime talking avatars to support agents, tutors, sales reps, companions, and training products, using your existing LLM, voice, data, and compliance stack.
Realtime
WebRTC streaming
One-shot
single reference image
BYO stack
your LLM · TTS · STT
Per-second
no minimums or seats
◇ Who you are
Different reader, different page.
Product / engineering
Add a realtime avatar to your product
Drop a passthrough WebRTC session into your app. Bring your own LLM, TTS, and audio. Hosted React SDK and REST API.
Read onAI / platform team
Use Atlas as your face-rendering layer
Keep your existing orchestration, prompts, memory, moderation, and routing. Atlas receives audio and returns video.
Read onSecurity / IT / procurement
DPA, dedicated capacity, SSO
SSO via your IdP, sub-processor list, security questionnaire support, retention controls available on enterprise contracts.
Read onExecutive / business owner
Lower avatar costs, avoid platform lock-in
Pay-per-second rendering, no seat fees, no expiring credits, no avatar-studio lock-in. The visual layer for AI products you already own.
Read on◆ Your data stack stays yours
Atlas renders the face. You keep the brain.
By default Atlas receives audio and returns rendered video. Your application controls user data, prompts, LLM calls, TTS, moderation, memory, and business logic. We never see your prompts, customer records, or knowledge base.
- No customer PII required for rendering
- We do not use customer inputs to train models
- Inputs processed in-memory; outputs purged on a short window
- Bring your own storage on enterprise contracts
◆ Security & compliance
Built so security review can sign off.
What ships today. We respond to standard security questionnaires (CAIQ-Lite, SIG-Lite, or your own format) during procurement, sign DPAs, and walk through architecture on a call.
SSO via your IdP
SAML 2.0 SSO available on enterprise contracts (Okta, Azure AD, Google Workspace).
Encryption
TLS 1.2+ in transit. Storage volumes encrypted at rest. Per-tenant API keys with rotation.
DPA + sub-processors
DPA available on request. Sub-processor list maintained on the privacy page; notification on material additions.
Tenant isolation
API keys scope every request. Dedicated GPU pools available on enterprise; no shared inference state.
Operational logging
Session create / delete and admin actions are logged for billing and incident review. Per-tenant log export sized per contract.
Data residency
The hosted service currently operates in the US. Discuss regional requirements before deployment.
Acceptable use & abuse response
Acceptable use policy, public-figure restrictions, session logging, and our abuse takedown path live on /safety.
◆ Reliability
What you get when something goes wrong.
99.9% uptime SLA
Available on enterprise contracts with dedicated GPU capacity. Credits on miss.
Incident communication
Direct customer notifications and severity-scoped response targets when service is disrupted.
Named technical contact
One owner on our side per enterprise account. Shared Slack/Teams channel.
Incident response
Direct ping on incident, severity-scoped response targets in your contract.