About Eleven v4 Dialogue
Eleven v4 Dialogue generates a single audio clip of several voices in conversation, with natural turn-taking, interruptions, and emotional delivery across more than 90 languages.
Each line pairs a piece of text with a voice, and the model performs the exchange as one scene rather than stitching separate clips, so timing and reactions carry across speakers. Audio tags such as [laughing], [sarcastic], or [interrupting] direct the delivery line by line. It suits podcast-style conversations, audio drama, game and animation scenes, and training data for conversational speech. A request can use up to 10 distinct voices and 10,000 characters of text in total.
| Metric | Value |
|---|---|
| Context Length | 10,000 characters, 10 voices |
| Multilingual | Yes, 90+ languages |
Ready to build with Eleven v4 Dialogue?
Try Eleven v4 Dialogue in the Workbench to prompt it, compare outputs, and iterate on prompts without writing any code. When you're ready to ship, call the same model from our API and build your own apps on top of it.
curl -sSf -X POST https://hub.oxen.ai/api/ai/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $OXEN_API_KEY" \
-d '{
"model": "elevenlabs-eleven-v4-dialogue",
"messages": [
{
"role": "user",
"content": "Try sending a message."
}
]
}'Request parameters
Every field Eleven v4 Dialogue accepts in the request body. See the API reference for response formats.
https://hub.oxen.ai/api/ai/chat/completions| Field | Type | Required | Default | Description |
|---|---|---|---|---|
inputs | array<object> | Yes | — | The lines of the conversation in order, each spoken by its own voice. Up to 10 distinct voices and 10,000 characters in total. Audio tags in square brackets direct each line's delivery. |
language_code | stringnullable | No | — | Optional ISO 639-1 language code (e.g. en, es, ja) to force the language and its text normalization. Leave empty to detect it from the text. |
stability | number | No | 0.5 | Lower values give more varied, expressive delivery; higher values keep the performance consistent across generations. Range: 0 – 1 |
similarity_boost | number | No | 0.75 | How closely each line follows its voice. Higher values match it more closely at some cost to naturalness. Range: 0 – 1 |
seed | integernullable | No | — | Optional seed for best-effort deterministic output. Range: 0 – 4294967295 |