Voxtral Transcribe 2

Key Features

• •Speaker Diarization: Generate transcriptions with speaker labels and precise start/end times. This feature is ideal for meeting transcription, interview analysis, and multi-party call processing, enabling clear attribution of who said what and when, even in conversations with multiple participants.
• •Context Biasing: Provide up to 100 words or phrases to guide the model toward correct spellings of names, technical terms, or domain-specific vocabulary. This improves accuracy on proper nouns and industry terminology that standard models often miss, particularly useful in healthcare, legal, or technical domains.
• •Word-Level Timestamps: Get precise start and end timestamps for each word, enabling applications like subtitle generation, audio search, and content alignment. Developers can index speech content at the word level for fine-grained analysis and synchronization.
• •Ultra-Low Latency Streaming: Voxtral Realtime delivers live transcription with configurable delay down to sub-200ms, enabling voice agents and real-time applications that require immediate feedback. It uses a novel streaming architecture that transcribes audio as it arrives, maintaining near-offline accuracy.
• •Multilingual Support: Transcribe in 13 languages including English, Chinese, Hindi, Spanish, Arabic, French, Portuguese, Russian, German, Japanese, Korean, Italian, and Dutch. Non-English performance significantly outpaces competitors, making it suitable for global deployment.
• •Noise Robustness: Maintain transcription accuracy in challenging acoustic environments such as factory floors, busy call centers, and field recordings. This ensures reliable performance in real-world conditions where background noise is a factor.

Use Cases

• •Meeting Intelligence: Transcribe multilingual meeting recordings with speaker diarization to attribute comments to participants. This provides clear, searchable transcripts at low cost, enabling teams to review discussions and extract action items efficiently.
• •Voice Agents: Build conversational AI agents with sub-200ms transcription latency for natural, responsive interactions. Connect Voxtral Realtime to an LLM and TTS pipeline to create voice interfaces that feel fluid and immediate.
• •Contact Center Automation: Transcribe customer calls in real time to analyze sentiment, suggest agent responses, and automatically populate CRM fields. Speaker diarization ensures clear attribution between agent and customer, improving post-call analytics.
• •Media and Broadcast: Generate live multilingual subtitles for broadcasts using Voxtral Realtime with context biasing for proper nouns and technical terms. This enables accurate, low-latency captioning for global audiences.
• •Compliance and Documentation: Monitor and transcribe interactions for regulatory compliance with full speaker attribution and timestamps. This creates precise audit trails for financial, legal, or healthcare environments, ensuring adherence to standards.

Who is this for?

Software developers building voice-enabled applications, AI engineers working on conversational agents, and product teams integrating speech-to-text. Contact center operators seeking real-time call transcription with analytics. Media and broadcast engineers needing live subtitling. Compliance officers in regulated industries requiring accurate speaker-attributed transcripts. Researchers deploying open-weights models on edge devices for privacy. Also suitable for meeting intelligence platforms, interview analysis tools, and multilingual documentation systems.