Give your AI agent a photorealistic face with Cara-4.1 Express. Anam’s full pipeline measured 1.24 s median reply latency and 1.42 s median connection time in our benchmark.

Try interactive avatar

See API

before

[0]

Voice chat

After

[1]

video call

before

[0]

Voice chat

After

[1]

video call

before

[0]

Voice chat

After

[1]

video call

Trusted by over 8,000 builders

What is an interactive avatar?

Definition

An interactive avatar is an AI character you talk to over live video. It listens to the user, works out a reply with a language model, and speaks it with a synthesized voice while a video model animates the face to match.

Versus pre-rendered video

AI video generators render a scripted clip ahead of time, and every viewer sees the same file. An interactive avatar has no script. Each reply is generated during the conversation, so it can answer follow-up questions, change topic, and stop when the user interrupts.

Why latency matters

Less waiting between turns keeps a conversation moving. In our full-pipeline benchmark, Anam had the fastest median replies at 1.24 s, measured from the end of your sentence to the start of the avatar’s reply. Median connection time was 1.42 s, from pressing Start to the first visible avatar frame.

Interactive avatar

An interactive avatar is a digital character that listens to a user and responds in real time, using speech, language model reasoning, and live animation to support a two-way conversation instead of a fixed script.

Sub-second latency

Sub-second latency is the end-to-end response time under one second between a user speaking and an avatar replying, covering speech recognition, model reasoning, speech synthesis, animation, and network streaming.

How a conversation works

1. Listen

The user’s audio streams to Anam over WebRTC. Speech recognition transcribes it, and voice activity detection works out when they’ve finished speaking.

2. Think

The transcript goes to the language model you pick, along with your system prompt, knowledge base and tools. Use Anam’s hosted models or bring your own LLM.

3. Speak

Text-to-speech turns the reply into audio. Pick a stock voice, clone your own, or connect a provider such as ElevenLabs or Cartesia.

4. Render

Cara-4.1 Express, our avatar model, generates each frame of the face at 25 fps in sync with the audio and streams it back as live video.

From Lab to production in ten lines

Read docs

JavaScript

Python

LiveKit (Python)

LiveKit (Node)

Pipecat

1import { createClient } from "@anam-ai/js-sdk";
2
3const anamClient = createClient(sessionToken, {
4    disableInputAudio: false,      // Set true to disable microphone
5    audioDeviceId: "device-id",    // Specify audio input device
6    voiceDetection: {
7        endOfSpeechSensitivity: 0.5, // 0-1, higher = more sensitive
8    },
9});

(Py)

(Node)

Pipecat

1import { createClient } from
2    "@anam-ai/js-sdk";
3
4const anamClient = createClient(
5    sessionToken, {
6    disableInputAudio: false,      //
7        Set true to disable microphone
8    audioDeviceId: "device-id",    //
9        Specify audio input device
10    voiceDetection: {
11        endOfSpeechSensitivity: 0.5, //
12            0-1, higher = more sensitive
13    },
14});

Video Agent Quickstart

Get started

Explore starter Projects

Get started

Integrate with LiveKit

Get started

Integrate with Pipecat

Get started

Fastest replies in our benchmark.
Less waiting between turns.

0.6s

0.6 s

faster replies than the nearest competitor

1.24s

1.24 s

median reply latency

Interruptible

If the user cuts in, the avatar stops talking and listens, the way a person would.

Support for 70+ languages

Run the same avatar in 70+ languages with native voices, covering 95% of the world’s population.

Benchmarked in public

Our October 16, 2026 study measured providers’ full pipelines with 60 raters and 960 conversations. See Avatar Benchmark, October 2026

Where teams use interactive avatars

Customer support

Answer product and account questions on your website or in your app, grounded in your knowledge base and connected to your own tools.

Sales training

Give reps and staff a realistic person to practise with, such as a skeptical buyer or an upset customer, as often as they need.

Language tutoring

Run spoken lessons and conversation practice in 70+ languages, with a tutor that responds to each learner’s answers.

Recruitment

Run the same first-round interview for every candidate, at any hour, and share the transcript with your hiring team. See recruitment avatars

Drops into the stack you already use

Logos for LiveKit, Pipecat, ElevenLabs, OpenAI, Deepgram, Cartesia, Google

Built for regulated buyers

Enterprise API

Enterprise-level data protection

SOC 2 Type II report and HIPAA-aligned controls, with end-to-end encryption and role-based access controls.

Zero Data Retention

Enterprise support and 99.9% uptime

Enterprise security and compliance icons

Built for regulated buyers

Enterprise API

Enterprise-level data protection

SOC 2 Type II report and HIPAA-aligned controls, with end-to-end encryption and role-based access controls.

Zero Data Retention

Enterprise support and 99.9% uptime

Check out interactive avatars

How Docket Increased Website Conversations 3.4x Using Anam’s Interactive Avatars

Read story

How Imeld Achieved 100% User Adoption Using Anam's Interactive Avatars to Create Video Agents

Read story

Anam’s Interactive Avatars Accelerated Airspeed’s Sales Cycle by 20%

Read story

How Anam’s Interactive Avatars Increased User Engagement 2X for ReplicateLabs’ AI Performance Coach

Read story

How Colleva Accelerates Time to Value by 10x and Reduces Hiring Costs by 70% Using Anam's Interactive Avatars

Read story

Read our research

0123456789

better overall experience than the nearest competitor

Overall experience

Lip sync

Visual quality

Naturalness

Responsiveness

Interruptibility

Anam

72%

Tavus

54%

HeyGen

47%

LemonSlice

49%

Share of ratings above neutral. Headline improvements compare the rounded positive-rating shares with the nearest competitor for each measure. Avatar Benchmark, October 2026.

Read our research

0123456789

better overall experience than the nearest competitor

Overall experience

Lip sync

Visual quality

Naturalness

Responsiveness

Interruptibility

Anam

72%

Tavus

54%

HeyGen

47%

LemonSlice

49%

Share of ratings above neutral. Headline improvements compare the rounded positive-rating shares with the nearest competitor for each measure. Avatar Benchmark, October 2026.

Frequently asked questions

What is an interactive AI avatar?

An AI character that holds a live, two-way video conversation. Anam’s avatars listen to the user, generate a reply with a language model, and speak it with a face rendered in real time. Try one in the AI avatar playground.

How is it different from an AI video generator?

A video generator turns a script into a clip before anyone watches it. An interactive avatar generates each reply during the conversation, so users can ask their own questions and get an answer straight away.

How fast does the avatar respond?

Cara-4.1 Express measured 1.24 s median reply latency with Anam’s full pipeline, from the end of your sentence to the start of the avatar’s reply. Median connection time was 1.42 s, from pressing Start to the first visible avatar frame. Results are from October 16, 2026; performance varies with your integration. See Avatar Benchmark, October 2026

Can I bring my own LLM and voice?

Yes. Use Anam’s hosted models or connect your own LLM, choose a stock, cloned or third-party voice, and run the avatar inside LiveKit, Pipecat, ElevenLabs Agents, Agora or VideoSDK.

Which languages are supported?

More than 70 languages, each with native voices. That covers about 95% of the world’s population.

How do I add an avatar to my website or app?

On a website, embed the Anam widget. There are setup guides for WordPress, Shopify, Wix, Squarespace and Framer. In an app, use the JavaScript or Python SDK, or call the REST API directly.

Can I create a custom avatar?

Yes. Describe the avatar in text, which works on every plan, or upload a photo and Cara-4.1 Express turns it into an animated avatar that can hold a conversation.

Can users interrupt the avatar?

Yes. Voice activity detection notices when the user starts speaking, and the avatar stops to listen.

Is it suitable for regulated industries?

Anam has a SOC 2 Type II report and HIPAA-aligned controls. Enterprise plans add zero data retention, so conversation data is not stored or used for model training.

Do I need to write code?

No. You can build and test an avatar in Anam Lab without code and embed it with the widget. Developers can use the SDKs and API for full control.

How much does it cost?

You can start free with 30 minutes a month. Paid plans are billed by the second, and the rate drops to $0.04 a minute at scale. See pricing

Try it out
for free.

For people that want to try out our playground

/month

Unlimited custom avatars

Text-to-Avatar

30 free minutes monthly

3 min conversation limit

Large scale needs?
Discover our enterprise plans.

Enterprise

Plans

For those wanting the full experience

Unlimited custom avatars

Uptime SLA

Dedicated support & SLA

Region Selection & Data Residency

Custom simultaneous sessions

Zero Data Retention

SOC 2 Type II report · HIPAA-aligned controls.

Interactive AI avatarsthat talk back in real time

What is an interactive avatar?

Definition

Versus pre-rendered video

Why latency matters

Interactive avatar

Sub-second latency

How a conversation works

1. Listen

2. Think

3. Speak

4. Render

From Lab to production in ten lines

From Lab to production in ten lines

Fastest replies in our benchmark. Less waiting between turns.

Where teams use interactive avatars

Customer support

Sales training

Language tutoring

Recruitment

Drops into the stack you already use

Built for regulated buyers

Built for regulated buyers

Frequently asked questions

Try it out for free.

Large scale needs? Discover our enterprise plans.

Enterprise

Plans

Interactive AI avatars
that talk back in real time

Fastest replies in our benchmark.
Less waiting between turns.

Try it out
for free.

Large scale needs?
Discover our enterprise plans.