# Real-time AI avatar

> A real-time AI avatar of your character: their face talks at 25 fps, lip-synced to their voice, as you speak with them. In the browser, or in your app by API.

_Updated 2026-11-21_

eroq Live turns the picture of one of your characters into a real-time avatar: as you talk, they answer in their voice and their face moves with every word, 25 frames a second. Call them in your browser, or put the same call in your own app through the API.

## At a glance

| | |
| --- | --- |
| Face | 512×512 at 25 frames a second, lip-synced to their voice |
| Voice | Theirs: the voice on their character card, cloned or from the library |
| Mind | eroq RP+, eroq's own roleplay model, as their persona |
| Languages | They answer in the language you speak |
| Price | 100 credits a minute, bought a minute at a time · nothing if the call never connects |
| API | POST /v1/live/sessions · WebRTC · billed per minute |

## One picture, a face that talks

- The face comes from the picture on the character's card: a clear, front-facing portrait works best.
- 512×512 at 25 frames a second, the lips in sync with the voice.
- Between replies the face stays alive: it blinks and moves while it listens.

## A mind and a voice behind it

- They think with eroq RP+, eroq's own roleplay model, as the persona on their card.
- They speak with their own voice, in the language you speak.
- They hear you through your microphone; your camera is never used.

## In your app

POST /v1/live/sessions opens a call with one of your characters; your page connects over WebRTC and buys the next minute as the call goes on, 100 credits a minute. A call that never connects gives its minute back.

## FAQ

### What is a real-time AI avatar?

A face that talks as the conversation happens: it hears you, answers in a voice, and its lips and expressions follow the words live, instead of a video rendered ahead from a script.

### Can I make an avatar of my own character?

Yes: any of your characters with a picture on their card. Make one from photos, a description or traits, give them a persona and a voice, then call them. A real, identifiable person needs their consent.

### Is it free?

Free to start: every new account gets 40 free credits with no card. A call costs 100 credits a minute, bought a minute at a time: the $10 pack (1,000 credits) is 10 minutes of calls, and a call that never connects costs nothing.

### Can the avatar read a script I write?

Not on a call: they answer what you say, in their own words, as their persona. Live is a conversation, not a video made from a script.

### Is there an API?

Yes: POST /v1/live/sessions with one of your characters returns the connection details, your page sends its WebRTC offer and plays the face and voice that come back, and /extend buys the next minute. Every call bills the same credits as the studio.

### Is the call recorded?

No. eroq keeps no audio or video of a call: the sound and the picture stream to your browser and are not saved. The call's record holds its length and the credits it used.

## Related

- https://eroq.ai/tools/ai-video-call — AI video call
- https://eroq.ai/docs/live — Live calls API
- https://eroq.ai/glossary/talking-head — Talking head
- https://eroq.ai/tools/ai-companion-video-call — AI companion video call
- https://eroq.ai/use-cases/ai-companion-apps — AI companion apps

---

This page as HTML: https://eroq.ai/tools/real-time-ai-avatar · Studio: https://eroq.ai/studio · Docs: https://eroq.ai/docs · Machine index: https://eroq.ai/llms.txt
