Whisper Large v3 Turbo
Streamlined open-source speech recognition model delivering near-identical transcription accuracy at a fraction of compute latency.
Best for: Real-time speech-to-text, video subtitle generation, and voice assistant frontends.
Specifications
- Provider: OpenAI
- Release date: October 2024
- Context window: 30s sliding audio chunk
- Modalities: Audio, Speech
- License: MIT Open Source
- Pricing: Open weights MIT; API $0.006 / minute
- API available: Yes
Capabilities
- 8x faster inference than v3
- 99+ language transcription
- Accurate timestamp generation
- Robust noise resilience