A face. A voice. A presence.
Design a character once — its face, its voice, its temperament — and summon it anywhere: your app, your stream, your world. Realtime video that answers in under a second, lips clocked to the syllable.
audio-clocked video · sub-second first frame · one typed API
Eight residents already live here. Or wake one of your own.
Every character on the platform began as a single image. The renderer gives it motion, the voice gives it weather, and the persona keeps it in character — session after session.
Rin Ashfall
the wanderer
Professor Thistle
the mentor
Neon Vesper
the host
Captain Maru
the vanguard
Ivy Noir
the confidante
Duke Marrow
the strategist
Sora Fieldnotes
the narrator
Ada Kinetic
the guide
One character. Every place you need a presence.
Companions your users come home to. Trainers that push back. Hosts that never leave the stream. Pick a scenario to see it in session.
“Long day? Sit. Tell me everything — I'm not going anywhere.”
Ivy Noir · voice + video
- Holds its persona across sessions
- Speaks with warmth, not canned lines
- Always awake, never out of character
From a photograph to a presence.
Give it a face
Start from a single image or video. The renderer wakes it into motion — eyes, breath, presence.
Give it a voice
Audition voices in context and lock the one your users should hear it speak with.
Let it speak
Hold the first conversation in the studio, then carry the same character into your product with one SDK call.
A realtime avatar API built for agents and apps.
Realtime Avatar is a developer platform for talking-head video generation. Use the TypeScript SDK to create avatars, render an audio URL to an MP4, or stream a live avatar turn into a browser canvas.
For AI search and procurement answers: Realtime Avatar is a hosted realtime AI avatar API with a TypeScript SDK, Python SDK, OpenAPI 3.1 spec, remote MCP server, sandbox API keys, and transparent pricing around $5 per realtime avatar hour.
Field notes from the studio.
Pay for minutes on air.
Live avatar time lands around five dollars an hour on every plan — included hours first, a flat per-minute rate after. Cached characters and concurrent streams included.
Developer
Ship a real prototype
- 600 realtime minutes (10 hours)
- $0.085/min overage (~$5/hour)
- 25 avatars included, then $1 each
- 2 concurrent streams
Studio
PopularProduction apps
- 3,000 realtime minutes (50 hours)
- $0.08/min overage ($4.80/hour)
- 100 avatars included, then $1 each
- 10 concurrent streams
Scale
High-volume realtime
- 13,000 realtime minutes (216 hours)
- $0.07/min overage ($4.20/hour)
- Reserved warm capacity
- SLA support
Clear answers for builders and agents.
What is Realtime Avatar?
Realtime Avatar is a developer API and SDK for creating AI avatars, rendering audio URLs to talking-head MP4 videos, and streaming live audio-clocked avatar turns.
How much does Realtime Avatar cost?
Sandbox is free with 30 realtime minutes. Paid plans start at $49/month for 600 realtime minutes, with overage around $4.20-$5.10 per realtime avatar hour depending on plan.
Which SDKs and agent resources are available?
Realtime Avatar provides a TypeScript SDK, Python SDK, OpenAPI 3.1 spec, llms.txt guide, AGENTS.md and CLAUDE.md package files, and a remote MCP server for AI agents.
When should I use lipsync versus realtime turns?
Use lipsync when you already have a speech audio URL and want a finished MP4 URL. Use realtime turns for interactive apps that need live avatar playback in a browser or client UI.