Voxtral Transcribe 2

Additional capabilities include speaker diarization that generates transcriptions with speaker labels and precise start/end times for meeting transcription and interview analysis, though with overlapping speech the model typically transcribes one speaker. Context biasing allows users to provide up to 100 words or phrases to guide the model toward correct spellings of names, technical terms, or domain-specific vocabulary, particularly useful for proper nouns and industry terminology. Word-level timestamps enable precise start and end timestamps for each word, supporting applications like subtitle generation and audio search, while expanded language support covers 13 languages with non-English performance significantly outpacing competitors. Longer audio support processes recordings up to three hours in a single request.

The product works through two specialized models with distinct technical approaches: Voxtral Realtime employs a streaming architecture purpose-built for live applications where latency matters, transcribing audio as it arrives rather than using chunk-based adaptations of offline models. Voxtral Mini Transcribe V2 utilizes batch processing optimized for accuracy and efficiency, delivering state-of-the-art transcription with speaker diarization, context biasing, and word-level timestamps. Both models share multilingual capabilities across 13 languages and support privacy-first deployment through on-premise or private cloud setups, with Voxtral Realtime available as open-weights under Apache 2.0 license for edge deployment.

Benefits and measurable outcomes include industry-leading accuracy at the lowest price point, with Voxtral Mini Transcribe V2 achieving the lowest word error rate while costing only $0.003 per minute. Users gain the ability to process audio approximately three times faster than competing solutions while matching quality at one-fifth the cost, enabling large-scale transcription projects previously limited by budget constraints. The sub-200 millisecond latency of Voxtral Realtime unlocks new classes of voice-first applications, allowing developers to build responsive voice interfaces that feel natural and immediate rather than delayed or disjointed.

Concrete use cases include meeting intelligence applications that transcribe multilingual recordings with speaker diarization clearly attributing who said what and when, enabling annotation of large volumes of meeting content at industry-leading cost efficiency. Voice agents and virtual assistants benefit from connecting Voxtral Realtime to LLM and TTS pipelines for responsive conversational AI with sub-200ms transcription latency. Contact center automation utilizes real-time call transcription enabling AI systems to analyze sentiment, suggest responses, and populate CRM fields while conversations are still happening, with speaker diarization ensuring clear attribution between agents and customers.

Target users include developers building voice applications, enterprises requiring meeting transcription and contact center automation, media companies needing live multilingual subtitles, and organizations with compliance documentation requirements. Integrations work with existing LLM and TTS pipelines, while the tech stack supports API access, open-weights deployment via Hugging Face, and private cloud implementations. Pricing plans include Voxtral Mini Transcribe V2 at $0.003 per minute via API and Voxtral Realtime at $0.006 per minute via API, with both models supporting GDPR and HIPAA-compliant deployments through secure on-premise or private cloud setups for sensitive data handling.

The primary value of Voxtral Transcribe 2 lies in delivering state-of-the-art transcription quality with unprecedented efficiency, enabling both real-time and batch processing applications through specialized models optimized for their respective use cases. By combining industry-leading accuracy, multilingual support across 13 languages, configurable low latency, and privacy-first deployment options, the product addresses critical pain points in speech-to-text technology while maintaining the lowest price point in the market. This combination of performance, flexibility, and cost-effectiveness transforms how organizations implement voice workflows across diverse applications from live voice agents to large-scale meeting transcription projects.

Updated 2026-02-28