Realistic AI avatars for your voice agent
Give your agent a lifelike face that speaks its voice in real time, with natural lip sync and expressions. A few lines of code, $0.02 a minute, and no subscription.
$1 in free credits (50 minutes) · No subscription · Billed per second
Works with any voice, from any provider
- OpenAI
- ElevenLabs
- Azure
- Fish Audio
- Cartesia
- LiveKit
- Daily
- Pipecat
- WebRTC
- per minute, pay as you go
- $0.02
- billing, never rounded up to the minute
- Per second
- subscription, setup or monthly minimum
- $0
- per session, as many sessions as you need
- 3 hours
Supported by
- AWS
- Cloudflare
- PostHog
- NVIDIA
- Azure
- AssemblyAI
- Google for Startups
Talking Avatar is supported by AWS Activate, Cloudflare, PostHog, NVIDIA Inception, Microsoft for Startups, AssemblyAI, and Google for Startups.
Get started
Three ways to give your agent a face
Code it with the API, build it without code, or run a 3D avatar free in the browser.
Developers · API
Realistic Avatars
$0.02 per minute
Stream your agent's voice to a photorealistic avatar with one npm package, or our LiveKit Agents plugin for Python. Works with any TTS.
No code
Voice Agents
$0.06 per minute, AI and voice included
Write a prompt, pick a face and a voice, then share a link or embed the agent on your site.
Free · MIT
Open Source
$0 forever
A 3D talking avatar that runs entirely in the browser, with a female and a male avatar built in, or your own .glb. No server, no API key.
Avatars
Realistic AI avatars for every audience
A growing library of ready-to-use photorealistic avatars, professional or casual, hosted and streamed to your page. Watch four of them below, or talk to one live for two minutes. Every avatar streams in 16:9 at 25 frames per second.
CasualSam
Navy henley, home study
CasualMaya
Knit cardigan, living room
ProfessionalDaniel
Dark suit, office
ProfessionalClaire
Blazer, office








- More
Plenty of faces to choose from
A growing library of women and men from their twenties to their sixties, from many backgrounds, in offices, studios and homes. Switch avatars per session with one ID, or have your own face made into one.
Voice Agents · no code
Launch a complete AI voice agent, no code needed
Write a prompt, pick a face and a voice, then share a link or embed it on your site. The AI model, the voice and speech recognition are included, for $0.06 a minute.
- Templates for support, sales, interviews and more
- Natural turn-taking and interruptions
- Answers from your documents: knowledge bases
- Tools, webhooks and a transcript of every call
- Share link, website widget or iframe
- $1 of free credits to start

MayaYour order shipped this morning. Want me to text you the tracking link?
Features
Everything a face-to-face agent needs
Built for live conversation, not pre-rendered video.
Any voice, any provider
OpenAI, ElevenLabs, Azure, Fish Audio, Cartesia or your own TTS. The avatar follows the audio, so no text or timings are needed.
Built for live conversation
Stream audio as your agent generates it; the avatar starts speaking moments later. Interruptions take one call and tell you how much was heard.
Voice and video in sync
Lips, expressions and voice arrive together over WebRTC, so they never drift apart. No media servers to run.
Drop-in for LiveKit
Already on LiveKit? Our plugin for LiveKit Agents in Python adds the avatar in three lines, or use the SDK on Node.js. It joins your room as a participant, with its video and voice as tracks.
Secure by design
Your API key stays on your server. Your page only plays the stream, and audio is used only to animate the avatar.
Pay only for what you use
$0.02 a minute, billed per second, the same price at any volume. No tiers, no seats, no commitments.
Performance
Low latency, at any scale
Your agent's voice streams to the avatar as it's generated, and the avatar starts speaking it less than 250 ms after it arrives. Run as many sessions side by side as you need, at no extra cost.
<250ms
Avatar response time
Measured on our servers across production sessions (October 2026): the time from your agent's audio reaching the avatar to the avatar speaking it. The network between our servers and your caller adds to what they hear.
- Extra for concurrent sessions
- $0
- Live video, voice in sync
- 25 fps
For developers · Realistic Avatars
A talking avatar API, live in an afternoon
Talking Avatar Realistic Avatars is a hosted, paid service of ready-made photorealistic avatars, rendered on our servers and streamed to your page over WebRTC or into your LiveKit room (with a LiveKit Agents plugin for Python), for $0.02 a minute. One npm package for your server and your page. Its /server and /live clients handle streaming, playback, interruptions and reconnection.
- 1
Create an account. It comes with $1 of free credits, 50 minutes of avatar time. Create an API key in the dashboard, and top up or turn on auto top-up when you need more.
- 2
Start a session on your server. Pick an avatar and stream your agent's voice to it as it's generated.
- 3
Show it on your page. One component plays the avatar's video and voice, and reconnects by itself.
npm install @talkingavatar/sdkimport { LiveAvatar } from "@talkingavatar/sdk/server";
const avatar = await LiveAvatar.start({
apiKey: process.env.TALKING_AVATAR_API_KEY,
avatar: "your-avatar-id", // from your dashboard
});
// One route for your page
app.post("/avatar", async (req, res) => {
res.json(await avatar.handleView(req.body));
});
// Your agent's voice, as it streams
tts.on("audio", (pcm) => avatar.pushAudio(pcm));
// The user interrupts
const heardMs = await avatar.clear();import { LiveAvatarView } from "@talkingavatar/sdk/live";
const view = new LiveAvatarView({
container: document.getElementById("avatar"),
endpoint: "/avatar",
});
startButton.onclick = () => view.connect();# pip install livekit-plugins-talkingavatar
from livekit.plugins import talkingavatar
# In your agent's entrypoint, before session.start(...)
avatar = talkingavatar.AvatarSession(avatar_id="your-avatar-id")
await avatar.start(session, room=ctx.room)
# The avatar joins the room and speaks your agent's voice.import { LiveAvatar } from "@talkingavatar/sdk/server";
// The avatar joins your room with a publish-only token you mint.
const avatar = await LiveAvatar.start({
apiKey: process.env.TALKING_AVATAR_API_KEY,
avatar: "your-avatar-id", // from your dashboard
livekit: { url: LIVEKIT_URL, token: avatarToken },
});
tts.on("audio", (pcm) => avatar.pushAudio(pcm));Use cases
Wherever your agent talks to people

Voice agents
Put a face on your AI voice agent
Turn a voice-only agent into a face-to-face conversation your users trust.
- OpenAI Realtime, LiveKit Agents, Pipecat or your own stack
- Any TTS voice, any language your TTS speaks
- Interruptions handled in one call
Custom avatars
Your face, or your brand's
Turn a founder, a recruiter or a hired presenter into a realistic avatar that speaks your agent's voice live. Only with the consent of the person on camera.
- 1
Tell us your use case. A two-minute form in your dashboard.
- 2
Pay once. $199 per avatar, by payment link.
- 3
Send a short video and voice recording. With the signed consent of the person on camera.
- 4
Go live. Your avatar appears in your account, ready for your API key.
One-time fee
$199per custom avatar
- Built from a short video and voice recording
- Private to your account
- Sessions at the same $0.02 a minute
- Commercial use
You pay only after we reply.
Talking Avatar Open Source · free
A free 3D talking avatar, ready to use
Talking Avatar Open Source gives your app a talking 3D avatar, directly in your users' browsers: a female and a male avatar are built in, or bring any .glb. A 1 MB model listens to the audio and moves the mouth in real time, with natural blinks and brows. No server, no API key, no cost. It's a separate product from Realistic Avatars: the avatar is a 3D model rather than a photorealistic person, and nothing is hosted by us.
- A 3D talking avatar that runs entirely in the browser: no server, no API key
- Lip sync from any audio, with natural blinks and brows
- Female and male 3D avatars built in, or any .glb with Oculus visemes
- Live audio streams and clips, React component included
- MIT License: free for commercial and noncommercial use
npm install @interviewflowai/talking-avatar threeOpen-source SDK & npm setupBrowser API referenceFree talking avatar
FAQ
Questions, answered
Pricing, setup, voices and privacy. Anything else? The docs go deeper.
What is Talking Avatar?
Talking Avatar by InterviewFlowAI gives AI voice agents a face, in two versions. Talking Avatar Realistic Avatars is a hosted, paid service of ready-made photorealistic avatars, rendered on our servers and streamed to your page over WebRTC or into your LiveKit room (with a LiveKit Agents plugin for Python), for $0.02 a minute. Talking Avatar Open Source is a free, MIT-licensed 3D talking avatar for the browser: it renders a built-in female or male 3D avatar, or your own .glb, and animates its face from any audio (lip sync from a 1 MB model, plus natural blinks and brows), with no server, account or API key.
What's the difference between Realistic Avatars and Open Source?
Realistic Avatars are photorealistic people, rendered on our servers and streamed to your page as video. You need an account and an API key, and pay $0.02 a minute after your $1 of free credits. Open Source is a 3D talking avatar: it renders a .glb avatar that you supply and animates its mouth, blinks and brows with a 1 MB model that runs in your users' browsers: no server, account or API key, and free under the MIT License. Both follow your agent's audio, and neither includes an LLM or TTS.
Which avatars can I choose from?
A growing library of ready-made photorealistic avatars: women and men from their twenties to their sixties and from many backgrounds, in professional looks (suits and blazers in offices) and casual ones (knitwear and shirts in homes and studios). Every account can use all of them at the same $0.02 a minute and pick a different one for each session by its ID. For a face of your own, a custom avatar is $199, one time.
How much do Realistic Avatars cost?
$0.02 per minute of session, billed per second. There is no subscription, setup fee or minimum: new accounts start with $1 of free credits, you add more when you need them, and each session draws from your balance. $10 of credits is 500 minutes.
How do I get started?
Create an account, which comes with $1 of free credits (50 minutes), and create an API key. Then install the SDK, start a session on your server and show the avatar on your page. You can be live the same day, and top up or turn on auto top-up when the free credits run out.
Which voices and voice agents work?
Any. The avatar follows the audio, so it works with OpenAI, ElevenLabs, Azure, Fish Audio, Cartesia or any other TTS, and with voice agents built on OpenAI Realtime, LiveKit Agents, Pipecat or your own stack.
Can I build an AI voice agent without code?
Yes. Talking Avatar Voice Agents are set up in the dashboard: start from a template, write the prompt, add your documents as a knowledge base, pick an avatar and a voice, then share the agent's link or embed it on your website with one line. The AI model, speech recognition and voice are included, for $0.06 a minute billed per second.
Does it work with LiveKit?
Yes. Instead of streaming to your page, the avatar joins your LiveKit room as a participant, with its video and voice as tracks. On LiveKit Agents for Python, install our plugin (pip install livekit-plugins-talkingavatar) and start the avatar with your agent session in three lines; on Node.js, use the SDK. The avatar gets a publish-only token, so your LiveKit secrets stay with you.
Can users interrupt the avatar?
Yes. One call stops the avatar mid-sentence and tells you how much of the reply the user heard, so your agent's transcript stays accurate.
How long can a session last?
Each session can run for up to 3 hours. Start as many sessions as you need; you only pay for the minutes they're live.
What happens to my audio?
It's used only to animate the avatar, and isn't stored. Your API key stays on your server; your users' browsers only play the stream. With Talking Avatar Open Source, audio never leaves the browser at all.
Can I try it for free?
Yes. Every new account gets $1 of free credits, 50 minutes of Realistic Avatars, so you can build and test with the real API before you pay. Talking Avatar Open Source is free for commercial and noncommercial use under the MIT License: a 3D talking avatar that runs your own .glb model directly in the browser, with no server or API key.
Can I have an avatar of my own face or presenter?
Yes. A custom avatar is $199 per avatar, one time. Request it from your dashboard; we reply with a payment link, then you send a short video and voice recording with the signed consent of the person on camera. Sessions with it cost the same $0.02 a minute.
Can I use it in a commercial product?
Yes, both. Realistic Avatars include commercial use, and Talking Avatar Open Source is MIT licensed and free for commercial and noncommercial use. Keep the copyright and permission notice with copies or substantial portions of the software.
Give your voice agent a face today
Sign up and get $1 of free credits, 50 minutes of realistic avatar time. Then $0.02 a minute, billed per second, with no subscription.
Compare:TavusHeyGen LiveAvatarD-ID AgentsSimliEvery avatar API priceAlternatives guides