Scribe - A highly accurate speech-to-text model by ElevenLabs supporting 99 languages.

## Definition of Scribe Scribe is a speech-to-text (ASR) model developed by ElevenLabs, renowned for its high transcription accuracy across 99 languages. It processes real-world audio scenarios, providing structured JSON outputs with features like word-level timestamps, speaker separation, and audio event tagging (e.g., laughter). ## Language Support in Scribe Scribe supports 99 languages, including widely spoken languages like English and Italian, as well as traditionally underserved languages such as Serbian, Cantonese, and Malayalam. It achieves notably low word error rates (e.g., 98.7% for Italian, 96.7% for English). ## Benchmark Performance of Scribe Scribe outperforms leading models like Gemini 2.0 Flash, Whisper Large V3, and Deepgram Nova-3 in benchmarks such as FLEURS and Common Voice. It achieves industry-leading word error rates and significantly reduces errors for underserved languages (where competitor models often exceed 40% error rates). ## Features of Scribe - **Multilingual Support**: 99 languages with high accuracy. - **Word-Level Timestamps**: Enables precise audio alignment. - **Speaker Separation**: Supports up to 32 speakers. - **Audio Event Tagging**: Identifies non-speech events (e.g., laughter, applause). - **Structured Outputs**: JSON format with metadata. - **Real-World Adaptability**: Works well in noisy environments or fast-paced speech. ## Access Methods for Scribe - **Developers**: Integrate via the [Speech-to-Text API](https://www.elevenlabs.io/docs/api-reference/speech-to-text/convert). - **Creators/Enterprises**: Use the [ElevenLabs Dashboard](https://elevenlabs.io/app/home) for no-code transcription. - **Pricing**: $0.40 per hour of input audio (50% discount for the first six weeks). ## Use Cases for Scribe Scribe is ideal for: - Meeting summaries and conference transcription. - Movie subtitling and lyric transcription. - Podcast and call center audio processing. - Future real-time applications (e.g., live meeting transcription). ## Unique Advantages of Scribe - **Language Inclusivity**: Focus on underserved languages. - **Low Latency (Upcoming)**: Planned support for real-time use cases. - **Comprehensive Metadata**: Includes speaker IDs and non-speech events. - **Competitive Pricing**: Cost-effective for high-volume users. ## Scribe Documentation Resources - **Blog Announcement**: [Meet Scribe](https://elevenlabs.io/blog/meet-scribe). - **API Docs**: [Speech-to-Text API Reference](https://www.elevenlabs.io/docs/api-reference/speech-to-text/convert). - **Dashboard**: [ElevenLabs App](https://elevenlabs.io/app/home). ### Citation sources: - [Scribe](https://elevenlabs.io/blog/meet-scribe) - Official URL Updated: 2025-04-01