---
title: Migrate from Voxtral 2
sidebar_label: Migrate from Voxtral 2
sidebar_position: 8
description: "Move from the Voxtral 2 audio models to Voxtral 3: model names, breaking changes, and a migration checklist."
---

# Migrate from Voxtral 2

This page is for anyone who calls a Voxtral 2 audio model. It maps each model to its Voxtral 3 replacement, then lists what changes in speech to text and text to speech, and what to do about each change.

<SectionTab as="h2" sectionId="model-mapping">Model mapping</SectionTab>

Replace each Voxtral 2 model ID with the ID of its Voxtral 3 replacement. Each Voxtral 3 model also answers to the aliases in the last column.

| Voxtral 2 model | Voxtral 3 model | Model ID | Aliases |
| --- | --- | --- | --- |
| `voxtral-mini-2602` | Voxtral Mini Transcribe 3 | `voxtral-mini-transcribe-3-0` | `voxtral-mini-transcribe-3`, `voxtral-mini-transcribe-latest` |
| `voxtral-mini-transcribe-realtime-2602` | Voxtral Mini Transcribe Realtime 3 | `voxtral-mini-transcribe-realtime-3-0` | `voxtral-mini-transcribe-realtime-3`,  |
| `voxtral-small-2507` | Voxtral Small Transcribe 3 | `voxtral-small-transcribe-latest` | `voxtral-small-transcribe-3`, `voxtral-small-transcribe-latest` |
| `voxtral-mini-tts-2603` | Voxtral TTS 3 | `voxtral-mini-tts-3-0` | `voxtral-mini-tts-3`, `voxtral-mini-tts-latest` will point to Voxtral 3 within a month of GA. |

<SectionTab as="h2" sectionId="speech-to-text-changes">Speech to text changes</SectionTab>

The following table lists the changes that can affect a transcription integration, with breaking changes first.

| Change | Breaks existing code? | What to do |
| --- | --- | --- |
| Word timestamps are gone. `timestamp_granularities` accepts `segment` only, and `word` returns a `400` error. | Yes, if you request word timestamps. | Request segment timestamps. See [Timestamps](/studio/audio/speech-to-text/transcribe-files#timestamps). |
| A realtime session locks its settings when the first audio chunk arrives. A later `session.update` gets an `error` event. |  | Send every setting in the first `session.update`, before any audio. See [Realtime transcription](/studio/audio/speech-to-text/transcribe-in-realtime#configure). |
|  | Yes, if your display assumes that segments never overlap. | Keep both segments when they overlap. See [Overlapping segments](/studio/audio/speech-to-text/diarize-speakers#overlapping-segments). |
| `context_bias` entries can be phrases with spaces, such as a full product name, and free-form text works offline. | No. | See [Context bias](/studio/audio/speech-to-text/bias-context-and-vocabulary). |
| `context_bias` has no fixed number of entries. | No. | Keep the list short: the entries share the context window with the audio, and a request over the limit returns a `400` error. See [Limits and errors](/studio/audio/speech-to-text/bias-context-and-vocabulary#limits-and-errors). |
| Transcription covers more languages. | No. | Omit `language` to detect the language automatically, or set it when you know it. See [Languages](/studio/audio/specs-and-limits/languages). |
| New: realtime diarization, speaker references, and turn detection in realtime sessions. | No. | Opt in when you need them. See [Diarize speakers](/studio/audio/speech-to-text/diarize-speakers) and [Turn detection](/studio/audio/speech-to-text/turn-detection). |

<SectionTab as="h2" sectionId="text-to-speech-changes">Text to speech changes</SectionTab>

The following table lists the changes that can affect a speech generation integration.

| Change | Breaks existing code? | What to do |
| --- | --- | --- |
| Emotion moves from the voice to inline tags. With Voxtral 2, each preset speaker has one variant per emotion, such as `gb_jane_sad`. Voxtral 3 adds `[emotion=...]` tags in `input` for finer control. | No. The Voxtral 2 presets are still accepted. | To vary the emotion of one voice, use [emotion and pace tags](/studio/audio/text-to-speech/emotion-tags). |
| Voice cloning for a single request with `ref_audio`. | No. `ref_audio` is still accepted. | To reuse a cloned voice, save it once and pass its ID as `voice_id`. See [Voice cloning](/studio/audio/text-to-speech/voice-cloning). |
| New: Voxtral 3 voices, pace tags, custom pronunciation, ~~style instructions~~ (not available at launch), and A number >= 1.0 that controls how expressive the speech is. Higher values trade voice similarity for stronger delivery. The default value is set by the service.. | No. | Opt in when you need them. See [Voices](/studio/audio/text-to-speech/voices), [Custom pronunciation](/studio/audio/text-to-speech/custom-pronunciation), [Style instructions](/studio/audio/text-to-speech/style-instructions), and [Expressiveness](/studio/audio/text-to-speech/generate-speech#expressiveness). |

Voxtral TTS 3 also changes how numbers, dates, prices, and email addresses are read, with no change to your request. The following inputs compare an earlier model with Voxtral TTS 3:

| Input | Earlier model | Voxtral TTS 3 |
| --- | --- | --- |
| Thank you for calling Mistral! I understand you're having difficulty logging into AI Studio. Can I confirm your email address? I have here elena.vance82@yahoo.com, is that correct? | <audio controls><source src="/audio/voxtral-3/expressiveness/customer-support-earlier-model.m4a" type="audio/mp4"/></audio> | <audio controls><source src="/audio/voxtral-3/expressiveness/customer-support-voxtral-3.m4a" type="audio/mp4"/></audio> |
| Gosh that was quick, I've found a flight for your upcoming travel. How about the Emirates flight at 21:00pm on 10/10 returning 10/25 at 22:00pm for £1881? Say the word, and I'll book a window seat now. | <audio controls><source src="/audio/voxtral-3/expressiveness/personal-agent-earlier-model.m4a" type="audio/mp4"/></audio> | <audio controls><source src="/audio/voxtral-3/expressiveness/personal-agent-voxtral-3.m4a" type="audio/mp4"/></audio> |

<SectionTab as="h2" sectionId="deprecation-timeline">Deprecation timeline</SectionTab>

<SectionTab as="h2" sectionId="migration-checklist">Migration checklist</SectionTab>

Follow these steps to move an integration to Voxtral 3:

1. Replace each model ID with its Voxtral 3 ID. See [Model mapping](#model-mapping).
2. If you request word timestamps, switch to segment timestamps.
3. In realtime sessions, send every setting before the first audio chunk.
4. If you diarize, check that your display handles overlapping segments.
5. Review your `context_bias` entries: write multi-word terms with spaces.
6. Generate a few samples with your voices, and move emotion variants to emotion tags if you want one voice in several moods. See [Text to speech changes](#text-to-speech-changes).
7. Measure latency again from your application. See [Latency](/studio/audio/concepts/latency).

<SectionTab as="h2" sectionId="going-further">Going further</SectionTab>

<UsefullLinkContainer>
  <LinkCard href="/studio/audio/speech-to-text/overview" title="Speech to Text" />
  <LinkCard href="/studio/audio/text-to-speech/overview" title="Text to Speech" />
  <LinkCard href="/studio/audio/specs-and-limits/limits" title="Limits" />
</UsefullLinkContainer>