[Enhancement]: Support for ElevenLabs' STT Model (scribe_v1)

[cranberry3148]

What features would you like to see added?

Support for ElevenLabs's "scribe_v1" Speech-to-Text model as an additional STT provider option in LibreChat.

More details

ElevenLabs recently launched "scribe_v1", their state-of-the-art STT model with:

• Industry-leading accuracy (outperforms OpenAI's 4o by 16% and 4o mini by 71% in WER according to third-party benchmarks)
• Support for 99 languages with exceptional performance across different accents
• Advanced features including:
  • Word-level timestamps with millisecond precision
  • Speaker diarization (up to 10 distinct speakers)
  • Audio event tagging (laughter, applause, etc.)
  • 5x faster than real-time processing

The REST API endpoint (https://api.elevenlabs.io/v1/speech-to-text) is straightforward to implement with structured JSON output and supports various audio formats.

Pricing is competitive at $0.22-$0.40 per hour of audio (depending on volume), making it cost-effective for production use.

Integration would enhance LibreChat with a high-performance STT option, especially benefiting users working with multi-language content or requiring high accuracy transcription.

References:

• https://elevenlabs.io/docs/capabilities/speech-to-text
• https://elevenlabs.io/blog/meet-scribe
• https://elevenlabs.io/blog/scribe-comparison-to-openais-4o-speech-to-text-model

Which components are impacted by your request?

Endpoints

Pictures

No response

Code of Conduct

• I agree to follow this project's Code of Conduct