Multilingual AI avatars
for builders

Build a real-time avatar for conversations in multiple languages. Choose speech recognition, a multilingual voice and instructions that fit your audience.

Trusted by 8,000+ builders

Anika, multilingual Anam avatar

One avatar, multiple languages in real-time

Your avatar listens and responds in a live conversation. Configure transcription and spoken language separately, and test your selected voice with the languages your audience uses.

1

Pick

Pick an avatar, a multilingual voice and an input language. Tell the LLM which language to speak.

2

Open

Test it in the Anam Lab, or use Anam’s JavaScript or Python SDK to drop it into your app.

3

Switch

Update transcription language during a call with a data-channel message or the change_language tool.

Unmatched realism.
Conversations across languages.

Meet Anika, powered by Cara-4.1 Express. Try to speak in another language or ask about setup. Voice coverage and pronunciation vary by model and language.

Anika, multilingual Anam avatar
Anika, multilingual Anam avatar
import { createClient } from "@anam-ai/js-sdk";

const response = await fetch("/api/session-token", { method: "POST" });
if (!response.ok) throw new Error("Could not create session");
const { sessionToken } = await response.json();
const client = createClient(sessionToken);

const video = document.createElement("video");
video.id = "avatar";
video.autoplay = true;
video.playsInline = true;

const language = document.createElement("select");
language.setAttribute("aria-label", "Conversation language");
for (const [code, label] of [["en", "English"], ["es", "Spanish"], ["fr", "French"]]) {
  language.add(new Option(label, code));
}
language.disabled = true;

const end = document.createElement("button");
end.textContent = "End call";
end.onclick = () => client.stopStreaming();
document.body.append(video, language, end);

await client.streamToVideoElement("avatar");
language.disabled = false;
language.onchange = () => {
  client.sendDataMessage(JSON.stringify({
    message_type: "persona_config",
    data: { languageCode: language.value }
  }));
  client.addContext(
    `The user selected ${language.selectedOptions[0].text}. Respond in that language.`
  );
};
import { createClient } from "@anam-ai/js-sdk";

const response = await fetch("/api/session-token", { method: "POST" });
if (!response.ok) throw new Error("Could not create session");
const { sessionToken } = await response.json();
const client = createClient(sessionToken);

const video = document.createElement("video");
video.id = "avatar";
video.autoplay = true;
video.playsInline = true;

const language = document.createElement("select");
language.setAttribute("aria-label", "Conversation language");
for (const [code, label] of [["en", "English"], ["es", "Spanish"], ["fr", "French"]]) {
  language.add(new Option(label, code));
}
language.disabled = true;

const end = document.createElement("button");
end.textContent = "End call";
end.onclick = () => client.stopStreaming();
document.body.append(video, language, end);

await client.streamToVideoElement("avatar");
language.disabled = false;
language.onchange = () => {
  client.sendDataMessage(JSON.stringify({
    message_type: "persona_config",
    data: { languageCode: language.value }
  }));
  client.addContext(
    `The user selected ${language.selectedOptions[0].text}. Respond in that language.`
  );
};
import { createClient } from "@anam-ai/js-sdk";

const response = await fetch("/api/session-token", { method: "POST" });
if (!response.ok) throw new Error("Could not create session");
const { sessionToken } = await response.json();
const client = createClient(sessionToken);

const video = document.createElement("video");
video.id = "avatar";
video.autoplay = true;
video.playsInline = true;

const language = document.createElement("select");
language.setAttribute("aria-label", "Conversation language");
for (const [code, label] of [["en", "English"], ["es", "Spanish"], ["fr", "French"]]) {
  language.add(new Option(label, code));
}
language.disabled = true;

const end = document.createElement("button");
end.textContent = "End call";
end.onclick = () => client.stopStreaming();
document.body.append(video, language, end);

await client.streamToVideoElement("avatar");
language.disabled = false;
language.onchange = () => {
  client.sendDataMessage(JSON.stringify({
    message_type: "persona_config",
    data: { languageCode: language.value }
  }));
  client.addContext(
    `The user selected ${language.selectedOptions[0].text}. Respond in that language.`
  );
};

Works with your LLM and voice

Preply learners practising with interactive avatars
Preply learners practising with interactive avatars
Preply

44% higher retention for tutor-assigned scenarios

Preply uses Anam’s real-time avatars to give learners more opportunities to practice speaking and build confidence between lessons.

Built for your most important conversations.

SOC 2 Type II

Move from pilot to production with independently audited security controls, backed by Anam’s SOC 2 Type II report.

HIPAA

Bring a human face to patient education, intake and clinical training, with security and privacy controls that support HIPAA-compliant deployments.

GDPR

Build for European audiences with privacy safeguards, clear data-processing commitments and support for GDPR requirements.

Frequently asked questions

How do I set up a multilingual avatar?

Choose an avatar, a voice that supports your target languages and a multilingual LLM. Set languageCode for transcription and use the system prompt to specify the language the avatar should speak. Create session tokens on your server, then connect your browser with the JavaScript SDK.

Does every voice support every language?

No. Transcription coverage, voice-model coverage and LLM language ability are separate. Anam offers 70+ transcription languages, including beta languages, but that does not mean every voice supports all of them. Test the chosen voice and speech-to-text provider with your audience’s languages and accents.

Can I switch languages during a call?

Yes. Send a persona_config data-channel message with languageCode, or attach the change_language system tool. This changes transcription; also tell the LLM which language to speak. Updates are best-effort and may take effect at the next user turn. Your voice model must support the requested language.

Can I use a custom avatar and tune pronunciation?

You can use a supported custom avatar with your chosen multilingual voice. Avatar appearance does not determine language coverage. Preview names and specialist terms in each target language. Pronunciation dictionaries can be configured for supported sessions; the saved-avatar API and embedded player may not expose the same voice options.

Can I bring my own LLM or agent framework?

Yes. With a custom LLM, your application supplies the responses while Anam handles transcription, voice and avatar media. With frameworks such as LiveKit or Pipecat, your agent pipeline supplies STT, LLM and TTS, and Anam supplies the avatar media. Configure and test language support in the components your integration owns.

Can the avatar answer from my knowledge base?

Attach a knowledge tool to folders containing your documents and instruct the avatar to retrieve relevant evidence before answering. Test retrieval and answers in your target languages. Azure/OpenAI-based LLMs can acknowledge a lookup and continue with the answer in the same turn; this interleaving is not supported by Groq-based models.

How do I update languages and knowledge after launch?

Update the saved avatar’s voice, language and prompt, then publish it for new sessions. Update the documents in its connected knowledge folders and test retrieval again. Active sessions keep their starting configuration unless you send a supported mid-session update; changing a saved avatar does not reset an ongoing call.