ElevenLabs introduced Scribe, a speech-to-text model designed to transform how we transcribe audio with remarkable accuracy and versatility. It supports 99 languages, making it a practical tool for a wide range of applications, from meeting summaries to subtitles and song lyrics.

Here’s what makes Scribe stand out:

1. Unmatched Accuracy : Outperforms leading models like Gemini 2.0 Flash, Whisper Large V3, and Deepgram Nova-3 In FLEURS & Common Voice benchmark tests across 99 languages.
2. Broad Language Support : Achieves remarkably low word error rates, including 98.7% in Italian and 96.7% in English.
3. Underserved Languages : Dramatically reduces errors in languages like Serbian, Cantonese, and Malayalam, where competing models often struggle with error rates above 40%.

Scribe doesn’t just stop at transcription. It enriches the experience with features like:

1. Word-level timestamps for precise tracking.
2. Speaker diarization to identify who’s speaking.
3. Audio-event tagging to capture non-speech elements like laughter or pauses.

For developers, Scribe offers:

1. A Speech-to-Text API for structured JSON transcripts.
2. Support for real-time applications with a low-latency version coming soon.

For creators and businesses, Scribe is accessible directly through the ElevenLabs dashboard , where you can upload audio or video files and generate formatted transcripts with ease.

Scribe is not just about accuracy; it’s about accessibility and practicality, making advanced transcription capabilities available to a broader audience. Whether you’re building applications or simply need reliable transcriptions, Scribe is ready to deliver. Try it out on the ElevenLabs dashboard.