Whisper

Whisper is OpenAI's open-source automatic speech recognition (ASR) model, released in 2022 and widely considered the most accurate publicly available transcription system. Trained on 680,000 hours of multilingual audio data, it supports 100 languages with high accuracy — including many low-resource languages that commercial transcription services struggle with — and handles accented speech, technical jargon, and background noise better than most alternatives.

The model is released under the MIT license, meaning it can be used free of charge for any purpose, including commercial applications. Running Whisper locally requires no API fees — compute costs are limited to your own hardware or a cloud instance. The model comes in five sizes (tiny, base, small, medium, large) with different speed/accuracy trade-offs; the large-v3 model delivers the best accuracy and fits on most modern GPUs with 10GB+ VRAM. OpenAI also provides a hosted API at $0.006/minute, which is competitive with commercial transcription services.

The primary limitation is that Whisper processes recorded audio files, not real-time streams — there is no built-in live transcription capability. Community projects like Whisper Live and WhisperStream add real-time functionality, but require additional infrastructure. For applications requiring live captions (video calls, live events), cloud-based services like AssemblyAI or Deepgram are better choices. For batch transcription of recordings, podcasts, meetings, and interviews, Whisper provides the best accuracy-to-cost ratio available.