ElevenLabs Unveils Innovative Speech-to-Text Model: Revolutionizing Voice Recognition Technology
ElevenLabs, an innovative AI startup renowned for its audio-generation capabilities, has recently secured a substantial $180 million funding round, enhancing its valuation to $3.3 billion. The company is now venturing into the realm of speech-to-text technology with the launch of its first standalone model, Scribe.
Introducing ElevenLabs’ Scribe: A New Era in Speech-to-Text Technology
With its extensive library of voices, ElevenLabs has already supported numerous companies in delivering effective speech-to-text services. The introduction of Scribe marks the company’s intent to compete in the growing market of speech detection technologies, challenging established players such as Gladia, Speechmatics, AssemblyAI, Deepgram, and OpenAI’s Whisper.
Key Features of Scribe
The Scribe model is designed to support over 99 languages at launch, with a focus on delivering accuracy. Here are some notable highlights:
- Excellent Accuracy: More than 25 languages boast a word error rate of less than 5%, including:
- English (97% accuracy)
- French
- German
- Hindi
- Indonesian
- Japanese
- Kannada
- Malayalam
- Polish
- Portuguese
- Spanish
- Vietnamese
- Performance Benchmark: Scribe outperformed Google Gemini 2.0 Flash and Whisper Large V3 across various languages in FLEURS & Common Voice benchmark tests.
Innovation in Speech Detection
Previously, ElevenLabs developed a speech-to-text component for its conversational agent platform. However, Scribe represents the company’s first foray into standalone speech detection. In a recent interview with TechCrunch, CEO Mati Staniszewski emphasized the need for improved speech detection models, stating:
“We want to understand what’s being said by you in a conversation better. Many people say that speech-to-text is a solved problem. But for many languages, it is pretty bad. We think we can build better speech detection models because we have in-house teams to annotate data and give us quick feedback.”
Advanced Features
Scribe comes equipped with several advanced functionalities:
- Smart Speaker Diarization: Identifies speakers in conversations.
- Word-Level Timestamps: Provides accurate subtitles.
- Auto-Tagging: Recognizes sound events, such as audience laughter.
Moreover, the platform enables customers to transcribe video content seamlessly for subtitle and caption integration.
Future Developments and Pricing
Currently, Scribe supports only pre-recorded audio formats. However, ElevenLabs plans to release a low-latency, real-time version of the model soon to facilitate meeting transcriptions and voice note-taking.
The pricing for Scribe is set at $0.40 per hour of transcribed audio, which remains competitive within the market, though some competitors may offer lower rates with different feature sets.
For more information about ElevenLabs and its innovative technologies, visit their official website here.
