Imagine a world where speech-to-text transcription happens so fast, it feels like magic. That's the reality with Voxtral Transcribe 2, a groundbreaking leap in speech recognition technology. Today, we're thrilled to unveil not just one, but two next-generation models that redefine what's possible in transcription quality, speaker identification, and real-time responsiveness. Meet Voxtral Mini Transcribe V2 and Voxtral Realtime, a dynamic duo designed to transform how we interact with voice data.
But here's where it gets exciting: Voxtral Realtime isn't just another model—it's open-source under the Apache 2.0 license, empowering developers to build privacy-first applications right on the edge. And to make things even more hands-on, we're launching an audio playground in Mistral Studio (https://console.mistral.ai/build/audio/speech-to-text), where you can test Voxtral Transcribe 2 instantly, complete with diarization and timestamps.
Highlights That Will Blow Your Mind
Voxtral Mini Transcribe V2: This powerhouse delivers state-of-the-art transcription in 13 languages, complete with speaker diarization, context biasing, and word-level timestamps. Whether you're transcribing meetings, interviews, or multi-party calls, it ensures every word is captured with precision. And here’s the kicker: it achieves the lowest word error rate at the lowest price point in the industry—just $0.003 per minute. Talk about a game-changer!
Voxtral Realtime: Built for live applications, this model offers ultra-low latency, configurable down to sub-200ms. This isn’t just fast—it’s revolutionary, enabling voice agents and real-time transcription with near-offline accuracy. And because it’s open-weights, you can deploy it on edge devices for maximum privacy and security.
Best-in-class efficiency: Voxtral Mini Transcribe V2 doesn’t just outperform competitors like GPT-4o, Gemini 2.5 Flash, and Deepgram Nova in accuracy—it does so at a fraction of the cost. Plus, it processes audio 3x faster than ElevenLabs’ Scribe v2 while matching its quality.
And this is the part most people miss: Both models are natively multilingual, excelling in 13 languages, including English, Chinese, Hindi, Spanish, Arabic, and more. This isn’t just transcription—it’s a global solution.
Voxtral Realtime: Redefining Real-Time Transcription
What sets Voxtral Realtime apart? Its novel streaming architecture. Unlike traditional models that process audio in chunks, Realtime transcribes audio as it arrives, delivering transcriptions with delays as low as sub-200ms. This unlocks a new class of voice-first applications, from live subtitling to responsive voice agents.
Controversial question: Could this be the end of offline transcription models as we know them? Let us know what you think in the comments!
At a 2.4-second delay, Realtime matches the accuracy of Voxtral Mini Transcribe V2, our batch model. Even at 480ms, it stays within 1-2% word error rate, making it ideal for voice agents that demand near-offline precision. With a 4B parameter footprint, it runs efficiently on edge devices, ensuring privacy for sensitive deployments.
You can find the model weights on the Hugging Face Hub (https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602).
Voxtral Mini Transcribe V2: The Gold Standard in Batch Transcription
This model isn’t just an upgrade—it’s a transformation. With significant improvements in transcription and diarization across languages and domains, it delivers a 4% word error rate on the FLEURS benchmark at an unbeatable price. It outperforms industry giants in accuracy and speed, making it the go-to choice for high-volume transcription needs.
Key Features That Stand Out:
- Speaker Diarization: Transcribe with speaker labels and precise timestamps, perfect for meetings and interviews. (Note: Overlapping speech is transcribed for one speaker at a time.)
- Context Biasing: Guide the model with up to 100 words or phrases to ensure accurate transcription of names, technical terms, or industry jargon. (Optimized for English, with experimental support for other languages.)
- Word-Level Timestamps: Generate precise timestamps for every word, enabling applications like subtitle generation and audio search.
- Noise Robustness: Maintain accuracy in challenging environments, from factory floors to busy call centers.
- Longer Audio Support: Process recordings up to 3 hours in a single request.
Transforming Industries with Voxtral
Voxtral isn’t just a tool—it’s a catalyst for innovation across industries:
- Meeting Intelligence: Transcribe multilingual meetings with clear speaker attribution, at industry-leading cost efficiency.
- Voice Agents: Build conversational AI with sub-200ms latency, creating natural, responsive interfaces.
- Contact Center Automation: Analyze sentiment, suggest responses, and update CRM fields in real time, all while ensuring clear speaker attribution.
- Media & Broadcast: Generate live multilingual subtitles with minimal latency and accurate handling of proper nouns.
- Compliance & Documentation: Monitor interactions for regulatory compliance, with diarization and timestamps for precise audit trails.
Both models support GDPR and HIPAA-compliant deployments, ensuring secure on-premise or private cloud setups.
Get Started Today
Ready to experience the future of transcription? Voxtral Mini Transcribe V2 is available now via API at $0.003 per minute. Test it out in the Mistral Studio audio playground (https://console.mistral.ai/build/audio/speech-to-text) or in Le Chat (http://chat.mistral.ai/).
Voxtral Realtime is available via API at $0.006 per minute and as open weights on Hugging Face (https://huggingface.co/mistralai/Voxtral-Mini-3B-Realtime-2602).
Explore the full documentation on Mistral’s audio and transcription capabilities (https://docs.mistral.ai/capabilities/audio_transcription).
We’re Hiring!
Passionate about pushing the boundaries of speech AI? Join our team and help us put frontier models into the hands of developers worldwide. Apply now at (https://mistral.ai/careers).
Thought-provoking question to end with: As transcription becomes faster, more accurate, and more accessible, how will it reshape industries like healthcare, media, and customer service? Share your thoughts below—we’d love to hear your perspective!
