Multilingual transcription
Voxtral Small Transcribe 3 and Voxtral Mini Transcribe 3 support multilingual speech recognition across 50 languages, including 24 official EU languages and lower-resource Asian and Indic languages.
Voxtral 3
Voxtral 3 combines multilingual speech transcription, expressive text-to-speech, and a real-time voice agent SDK for natural phone and web conversations.
Overview
Voxtral 3 is Mistral’s family of speech-to-text and text-to-speech models together with a voice agent SDK. It is designed for real-time, multi-step conversations in which an agent listens, interprets requests, uses connected tools, and responds with speech.
The product supports transcription in 50 languages, expressive speech generation in more than 35 languages and 50+ dialects, and deployment through Mistral’s EU-hosted API or on an organization’s own infrastructure. Its intended applications include contact-center and service-desk agents, virtual assistants, and hands-free copilots.
Core capabilities
Diarization and contextual recognition
The transcription models identify which speaker said what, detect when a speaker has finished, and accept keyphrases or free-form context to improve recognition of domain-specific terms without fine-tuning.
Expressive text-to-speech
Voxtral Mini TTS 3 generates speech in more than 35 languages and 50+ dialects, adapting emotion, pace, and emphasis to conversational context.
Real-time voice agent orchestration
The voice agent SDK links audio input, model reasoning, tool calls, and spoken output in a real-time streaming pipeline for live interactions.
Tool-enabled workflows
Agents can retrieve information, interact with business systems, and carry out multi-step actions through connected tools during phone or web conversations.
Professional voice adaptation
Pro Voice adaptation can reproduce a speaker’s distinctive character, emotional range, and speaking style across languages using 30 minutes of audio, according to the product page.
Use cases
Contact centers and service desks
A live service agent can transcribe a caller, distinguish speakers, detect turn completion, retrieve account or case information, and complete multi-step actions through connected business tools.
Conversational virtual assistants
Organizations can build assistants that handle spoken requests, reason over context, and respond with speech that varies its pace, emphasis, and emotion during the exchange.
Hands-free copilots
Voice interfaces can let users operate workflows while driving, moving through a facility, or otherwise keeping their hands and eyes occupied, with tool calls handling connected tasks.
Domain-specific meeting and call capture
Teams can use multilingual transcription with context biasing to capture conversations containing technical, medical, operational, or other specialized terminology without fine-tuning the model.
Pros and Cons
Pros
- Combines transcription, reasoning, tool calling, and speech generation in a single real-time voice-agent workflow.
- Supports 50 transcription languages and more than 35 TTS languages with 50+ dialects.
- Includes diarization, end-of-turn detection, and context biasing for production-oriented speech recognition.
- Offers an EU-hosted API and self-deployment option for organizations that need control over audio processing and retention.
- Supports expressive TTS and professional voice adaptation across languages.
Cons
- The source does not provide complete setup, SDK documentation, platform compatibility, or integration details.
- Pricing is only specified for the Mini Transcribe model; costs for the broader model family and SDK require confirmation.
- Real-world latency, accuracy, retention behavior, and operational requirements will depend on the selected deployment and implementation.
FAQ
What is Voxtral 3?
Voxtral 3 is a family of speech-to-text and text-to-speech models paired with a voice agent SDK for real-time streaming conversations. The SDK combines listening, reasoning, tool calling, and speaking in one workflow.
Which languages does Voxtral 3 support?
The transcription models support 50 languages. Voxtral Mini TTS 3 supports more than 35 languages and 50+ dialects.
How can Voxtral 3 be deployed?
Voxtral 3 is available through Mistral’s EU-hosted API or for self-deployment on an organization’s own infrastructure. The source states that deployment choice can determine where audio is processed and whether it is retained.
What can a Voxtral voice agent do during a conversation?
The SDK is designed for live conversations over phone or web and can retrieve information, interact with business systems, and complete multi-step actions through connected tools.
How is Voxtral 3 priced?
The transcription API is priced per minute on Mistral’s pricing information. The product page specifically lists Voxtral Mini Transcribe 3 at $0.0035 per minute; pricing for the full Voxtral 3 family and SDK should be confirmed with Mistral.
Quick Facts
- Category
- Speech AI and voice agent SDK
- Transcription languages
- 50
- Text-to-speech coverage
- 35+ languages and 50+ dialects
- Deployment
- Mistral EU-hosted API or self-deployment
- Transcription pricing example
- Voxtral Mini Transcribe 3: $0.0035 per minute
- Primary workflows
- Phone and web conversations with tool-enabled actions
Voxtral 3 Alternatives
SpeakoFlow
SpeakoFlow is a free, open-source desktop voice-to-text app for Windows, macOS, and Linux. Dictate into any app, use Flow, cleanup, translation, and a screen-aware assistant.
Talkpal
Talkpal is an AI language learning app for web and mobile with speaking, listening, writing, pronunciation practice, guided courses, roleplays, and 130+ languages.
Flunkey
Flunkey is a voice-first productivity layer for Windows that turns speech into text, AI outputs, and remembered context without leaving your app.
PXZ AI
An All-In-One AI Platform that combines tools for image, video, voice, writing, and chat to enhance creativity and collaboration.
Gemma AI
Gemma AI is a phone call reminder app that calls you with scheduled reminders instead of push notifications, with Google Calendar sync and natural voice interaction.
CAMB.AI Streams
CAMB.AI Streams dubs live audio in real time for YouTube, Twitch, X and other platforms, using existing live workflows and no post-production.