HeyGen LiveAvatar: what it is, what it costs and what you still build
HeyGen LiveAvatar is HeyGen’s real-time avatar product: an API and SDK that streams a lip-synced avatar into a live conversation, so your software can talk with people face to face. It is the production version of HeyGen’s Interactive Avatar (the old streaming avatar API), runs on its own platform at liveavatar.com with its own credits, and costs $0 for a 10-credit test tier, $99 or $475 a month for the self-serve plans, or a custom enterprise quote.
A disclosure: Life Inside’s talking avatar, the AI Video Agent, is built on HeyGen’s real-time avatar technology, so we are not neutral. This page covers what changed from Interactive Avatar, how developers use LiveAvatar, what it costs, what you still have to build to put it on a website, and when a packaged option is simpler. More tools on our AI avatar tools page.
LiveAvatar plans and what you get (checked 18 November 2026)
| Plan | Price | Credits and minutes | Limits and extras |
|---|---|---|---|
| Sandbox | Free | No credits used | For development only: one test avatar, sessions end after about a minute. |
| Free | $0 | 10 credits a month: 5 minutes in FULL mode or 10 in LITE mode | 1 session at a time, 2 minutes per session, watermark, preset 1080p avatars, full API access. |
| Essential | $99/month | 1,100 credits: 550 FULL minutes or 1,100 LITE minutes. Overage $0.095 a credit | Unlimited concurrent sessions, 20 minutes per session, 1 custom avatar (720p), no watermark. |
| Business | $475/month | 6,000 credits: 3,000 FULL minutes or 6,000 LITE minutes. Overage $0.09 a credit | Unlimited concurrent sessions, 60 minutes per session, 1 custom avatar (1080p), no watermark. |
| Enterprise | Custom quote | Rates down to $0.01 a minute at the top volume tier | Custom session length, dedicated capacity and SLA, SSO, admin control, audit logs, avatar fine-tuning. |
Is HeyGen LiveAvatar right for you?
- Use LITE mode if you already run a voice agent, your own or one on LiveKit, Pipecat, ElevenLabs, OpenAI Realtime or Gemini Live, and want to give it a face. At 1 credit a minute it is the cheaper mode.
- Use FULL mode if you want LiveAvatar to run speech recognition, the language model and the voice, and you will build the product and the experience around it.
- Use a HeyGen video plan instead if you need recorded videos to publish, not live conversations. See HeyGen pricing.
- Use [Life Inside](/ai-video-agent) if you want a talking avatar that answers visitors on your website without building the knowledge base, widget, guardrails and analytics yourself. Its avatar is built on HeyGen.
HeyGen interactive avatar, now LiveAvatar: what changed
LiveAvatar started as Interactive Avatar, a beta feature in HeyGen Labs, often reached through the Streaming Avatar SDK. HeyGen’s help centre now calls LiveAvatar the official production version of Interactive Avatar, and HeyGen’s developer docs send real-time avatar work to the LiveAvatar API. If you search for the HeyGen streaming avatar today, this is the product you land on.
- A separate platform: its own dashboard, API and docs at liveavatar.com. You sign in with your HeyGen account, but the workspace is separate.
- Separate credits: HeyGen plan credits and HeyGen API credits do not transfer to LiveAvatar, and LiveAvatar has its own plans.
- Separate avatars: avatars made in the HeyGen video app are not compatible with LiveAvatar. A custom LiveAvatar is trained from about two minutes of uncut footage (15 seconds listening, 90 talking, 15 listening), or from a single photo.
- Two modes: FULL, where LiveAvatar runs the whole conversation, and LITE, where your own agent sends audio and LiveAvatar renders the face.
- Fairer metering: the session starts before the client token is generated, so you are not charged while your frontend is still setting up.
- Migration: HeyGen offers migration support to existing Interactive Avatar users. Check your old avatars and integrations, as not every asset moves across.
How the LiveAvatar API works for developers
LiveAvatar streams over WebRTC through LiveKit rooms. Your backend creates a session with your API key, your frontend joins the room, and the avatar speaks in it. LiveAvatar quotes a median time to first frame under 300 ms.
- Plugins and connectors: LiveKit, Pipecat, Agora and Vision Agents, plus connectors for ElevenLabs Agents, Cartesia, OpenAI Realtime and Gemini Live.
- Language model in FULL mode: managed OpenAI models (the default became gpt-5.4-nano in September 2026), or your own model through a custom LLM integration.
- Voices: text-to-speech from ElevenLabs, Fish Audio or Cartesia, including your own keys.
- Quick embed: one API call returns an iframe you can paste into a page, useful for a first test.
- Get an API key from the LiveAvatar dashboard and test in Sandbox mode, which uses no credits.
- Pick an avatar: one of 100+ presets, or your own custom avatar.
- Choose a mode. In FULL mode you give it a voice and a context: a prompt, an optional opening line and optional links it can draw on. In LITE mode you connect your own agent.
- Create a session token on your backend, then start the session. The response holds the LiveKit room details.
- Join the room from your page with the LiveAvatar Web SDK or a LiveKit client. In LITE mode your agent streams its audio to the session.
- Keep the session alive while the visitor talks, stop it when they leave, and fetch the transcript afterwards.
curl -X POST https://api.liveavatar.com/v1/sessions/token \
-H "X-API-KEY: $LIVEAVATAR_API_KEY" \
-H "content-type: application/json" \
-d '{"mode": "FULL", "avatar_id": "<avatar_id>",
"avatar_persona": {"voice_id": "<voice_id>", "context_id": "<context_id>", "language": "en"}}'LiveAvatar pricing: credits, minutes and the real per-minute rate
LiveAvatar charges per minute of session time. It checks all active sessions every minute and deducts credits by mode: 2 credits a minute in FULL mode, 1 in LITE mode. A session cannot start without credits for at least one minute. Without overage switched on, sessions end when the balance runs out; with it, each extra credit is billed at your plan’s overage rate.
- Essential, $99 a month: about $0.18 a FULL minute and $0.09 a LITE minute. Overage works out at $0.19 and $0.095.
- Business, $475 a month: about $0.16 a FULL minute and $0.08 a LITE minute. Overage works out at $0.18 and $0.09.
- Enterprise: rates down to $0.01 a minute at the top volume tier, on contract.
- Example: 1,000 conversations a month of 3 minutes each is 3,000 minutes. In FULL mode that is 6,000 credits, the Business plan at $475. In LITE mode it is 3,000 credits: Essential plus 1,900 overage credits, $279.50, plus the cost of your own language model and voice.
What you still build to put LiveAvatar on a website
LiveAvatar gives you a face, a voice and, in FULL mode, a managed conversation. A website agent that visitors trust needs more around it, and that work is yours:
- The knowledge: a context holds a prompt and some links. Answers grounded in your whole website, documents and FAQs need retrieval you build or a custom language model, and someone has to keep the content correct.
- The interface: a widget that fits your site, microphone permission, mobile layouts, and a text fallback for visitors who will not talk out loud.
- The backend: a server that creates session tokens without exposing your key, keeps sessions alive and ends them so idle tabs do not burn minutes.
- Guardrails: what the agent must not answer, how it hands over to a person, and how you test it before launch.
- Analytics: the API returns transcripts. Topics, unanswered questions, drop-offs and leads are a reporting layer you build.
- Integrations: booking links, CRM and alerts, through your own code.
- Privacy and disclosure: consent, a data processing agreement, and telling visitors they are talking to AI, as the EU AI Act requires.
The packaged option: Life Inside’s AI Video Agent
Life Inside’s AI Video Agent is built on HeyGen’s real-time avatar technology and delivered as a managed service, so the list above is done for you. Life Inside builds the knowledge base from your website, documents and FAQs, you add one JavaScript snippet, and most agents go live within a day.
The avatar is an add-on to the AI Chat Agent, which answers in text first, in 60+ languages, and can show interactive video of your real people in the same window. AgentLoop shows what visitors asked, where they dropped off and which answers are missing. A custom avatar of one of your own people needs about two minutes of footage, with their consent. Life Inside is GDPR compliant; the avatar itself runs through HeyGen in the USA. See pricing or book a demo.
When LiveAvatar itself is the better choice
Build on LiveAvatar directly if the avatar is part of your own product: a tutoring or coaching app, a training simulator, kiosk software or a companion app. There you need API control, your own pipeline and volume pricing, and a managed website agent would get in the way.
Life Inside is not a developer API, does not produce videos to publish, and does not run live video calls between people. If you are still choosing a real-time platform, compare the options in our guide to the best interactive AI avatars and what conversational avatars are.
Frequently asked questions
Sources
- LiveAvatar by HeyGen: product and pricing (Checked 2026-11-18)
- LiveAvatar docs: overview and quickstart (Checked 2026-11-18)
- LiveAvatar docs: credits and subscriptions (Checked 2026-11-18)
- LiveAvatar docs: LITE or FULL mode (Checked 2026-11-18)
- LiveAvatar docs: FULL mode overview (Checked 2026-11-18)
- LiveAvatar docs: contexts (Checked 2026-11-18)
- LiveAvatar docs: sandbox mode (Checked 2026-11-18)
- LiveAvatar docs: changelog (Checked 2026-11-18)
- HeyGen Help Center: introducing LiveAvatar (Checked 2026-11-18)
- HeyGen blog: LiveAvatar by HeyGen (Checked 2026-11-18)
- LiveAvatar: HeyGen vs LiveAvatar (Checked 2026-11-18)
- HeyGen developer docs: live avatar (realtime) (Checked 2026-11-18)
Updated 2026-11-18