01 / API
API
- ▸REST API
- ▸Async jobs
- ▸Signed webhooks
Submit audio and a reference image, track the job, and retrieve the completed video.
Realtime AI
avatars
that talk back.
Turn a reference photo into a real-time, lip-synced video agent your users can actually talk to. Photorealistic, realtime, and ready for your product.
Compatible with
Logos are customer-provided integrations. Atlas is not affiliated with or endorsed by these providers.
Our avatar
Curated outputs
See them come alive
A tighter reel of generated avatars so visitors can evaluate the visual quality before exploring the full output showcase.
Hover or tap to preview motion. Audio is optional.
Using ElevenLabs voices. Hover any avatar to preview motion.
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
Speaking…
AI professional
◆ Plugs into your stack
Your LLM.
Your voice.
Our face.
Atlas renders the avatar; you keep control of the brain, the voice, and where the video goes. See the API docs.
No fixed realtime session-length cap.
Active session time is billed by the second; capacity and request limits still apply.
Any LLM, via your API key
Your fine-tuneWe don't lock you into a model. Plug yours in and we render the face.
Any voice
CartesiaAzureYour ownAtlas follows the supplied audio; test the voices and languages required by your product.
Anywhere you stream
WebRTCBrowserNative appYour client
BrowserNative appYour clientStream an Atlas avatar inside your own product and user experience.
Logos are customer-provided integrations. Atlas is not affiliated with or endorsed by these providers.
Product surfaces
Generate clips,
or render avatars live.
Offline mode turns a face and audio into a polished MP4. Realtime mode accepts audio from your stack and returns lip-synced WebRTC video.
⌘ 00 / Deployment
Built to ship.
Built to scale.
Whether you're integrating via REST, streaming over WebRTC, or adding Atlas to a production workflow, use the interface that fits your application.
Atlas Realtime Avatar
4 surfaces · 1 model family
Get started →↗
02 / Realtime
Realtime
- ▸WebRTC streaming
- ▸Bring your own audio
- ▸No fixed session-length cap
Publish audio from your application and receive a lip-synced avatar stream from a single reference image.
03 / Enterprise
Enterprise
- ▸Dedicated capacity
- ▸Contracted throughput
- ▸SLA & support
For larger organizations adding the Atlas Realtime Avatar API to their products and services with dedicated capacity and support.
04 / Dashboard
Dashboard
- ▸API keys
- ▸Usage
- ▸Self-serve access
Manage access to Atlas and begin integrating the API from the hosted dashboard.
◆ Built for Enterprise
Enterprise controls and support.
SCIM, SSO, dedicated GPU capacity, passthrough by default, and a real audit trail. Enterprise contracts only.
SCIM / SAML SSO
Enterprise IAM via your IdP. Provision and de-provision users programmatically; no per-seat fees.
Passthrough by default
Atlas receives audio and returns video. Your prompts, customer data, and knowledge base never leave your stack.
Dedicated capacity + SLA
Reserved GPU capacity, contracted throughput, and uptime terms scoped in Enterprise agreements.
Audit log + security review
DPA, sub-processor list, and security questionnaire support for procurement.
◇ Stuff people ask
Real questions from real DMs.
I'm building this site myself, so this is the running list of the questions that actually come up, in DMs, intro calls, partner emails. If yours isn't here, ask me and I'll add it.
Q · 01
Who is Atlas for?
Anyone shipping a product where a face needs to talk, devs, founders, agencies, enterprise. Specifically not a consumer app: we're the engine other products plug into. If you want to log in and post short-form video to social, there are great studio products for that. If you want a talking face inside your product, you're in the right place.
Q · 02
What's actually included in the price?
Just the avatar, the face talking. You bring your own TTS (ElevenLabs, Cartesia, OpenAI TTS) and your own LLM (OpenAI, Anthropic, Mistral, or a fine-tune). It keeps you in control of voice, prompt, and cost.
Q · 03
What can I actually build with this?
Realtime support agents, sales SDRs, NPCs, language tutors, virtual companions, async sales outreach at scale, training videos, onboarding flows, news anchors. Rough rule: if a face talking to a person makes the experience better, you're in the right neighborhood.
Q · 04
Do you do custom integration?
Enterprise contracts can include scoped integration support. Send us your architecture, volume, and deployment requirements so we can confirm what belongs in a statement of work.
◇ One model family · Endless surfaces
Build a tutor.
One avatar foundation model. Every application that needs a face, a voice, and a presence, shipped on your stack.