Whisper

OpenAI's open-source speech recognition model that delivers near-human accuracy across 99 languages. Whisper runs locally or via API, handling accented speech, background noise, and technical jargon with impressive reliability — making it the foundation for countless transcription tools and voice applications.

• Multilingual — supports transcription in 99 languages with automatic language detection • High accuracy — approaches human-level performance even with noise and accents • Open-source — download and run locally on your own hardware at no cost • Translation — translate foreign language audio directly to English text • Timestamped output — word and segment-level timestamps for precise alignment • Multiple model sizes — choose from tiny to large models based on accuracy vs. speed needs

Completely free and open-source. Also available via OpenAI API.