Introducing Grok Voice Transcribe 2.0
The new model keeps pricing at $0.10 per hour for batch and $0.20 per hour for streaming while doubling accuracy over version 1.0.
- Automatically detects dozens of languages and handles mid-recording language switches in a single pass.
- Reduced word error rate from 20.6% to 6.8% on short multilingual voice command evaluations.
- Includes diarization, timestamps, and key terms in the base pricing.
- Becomes the default model in the Speech-to-Text API, with Grok Voice Transcribe 1.0 scheduled for deprecation in coming weeks.
Developers using xAI's speech API get improved transcription accuracy on noisy and multilingual audio with no pricing changes or code updates required.
