Speech generation

A speech request turns a complete text into an audio file in one call, in the voice and the format you choose. This page shows the request, the fields that change the result, and the errors you're most likely to meet.

Send the text and a voice in one POST request. The response carries the audio, base64-encoded, in audio_data: decode it and save it to a file.

import base64
import os
from pathlib import Path

from mistralai.client import Mistral

client = Mistral(api_key=os.environ["MISTRAL_API_KEY"])

response = client.audio.speech.complete(
    model="voxtral-mini-tts-3-0",
    input="Your order shipped this morning and should arrive on Thursday.",
    voice_id="gb_jane_neutral",
    response_format="mp3",
)

audio = base64.b64decode(response.audio_data)
output = Path("order_update.mp3")
output.write_bytes(audio)
print(f"Saved {output.resolve()} ({len(audio)} bytes)")

import base64
import os
from pathlib import Path

from mistralai.client import Mistral

client = Mistral(api_key=os.environ["MISTRAL_API_KEY"])

response = client.audio.speech.complete(
    model="voxtral-mini-tts-3-0",
    input="Your order shipped this morning and should arrive on Thursday.",
    voice_id="gb_jane_neutral",
    response_format="mp3",
)

audio = base64.b64decode(response.audio_data)
output = Path("order_update.mp3")
output.write_bytes(audio)
print(f"Saved {output.resolve()} ({len(audio)} bytes)")

The examples save order_update.mp3 in the current directory. Open it in any audio player to listen.

The following table lists the fields that most often change the result. For the full list, see the API reference.

Set response_format to mp3 (the default), wav, pcm, flac, or opus. To choose between them, see Audio formats.

expressiveness sets how strongly the generated speech follows what you ask for: the voice, the text, and any emotion tags or style instruction. It accepts values from 1.0 to 5.0.

Controls how expressive the speech is. Higher values trade voice similarity and stability for stronger delivery. Start with the default, raise it if the emotion doesn't come through, and lower it if the speech sounds forced.

These are the problems you're most likely to meet with a speech request. For the full list, see Troubleshooting.