Open speech recognition
Whisper is a speech recognition model OpenAI released in 2022. It converts speech to text with high accuracy across many languages. What mattered most was that it was released openly and free, and it became the standard component inside transcription tools and apps worldwide.
What made it notable
- Robust to noise and natural speech, having trained on a huge volume of real-world audio
- Multilingual, with strong accuracy across languages and support for translating while transcribing
- Openly published, so anyone can download it and build it into their own machine or service
Accurate transcription, which had required expensive specialist services, became broadly available almost overnight.
Where you encounter it
Even without knowing the name, you have likely used it — meeting-notes apps, automatic video captions, and voice memo transcription frequently run Whisper or a derivative underneath. Because it runs on your own machine, it is also the standard choice for transcribing sensitive audio without sending it anywhere, on the same reasoning as a local LLM.
Worth knowing
Knowing it as the representative speech recognition model, widely adopted because it is open, is enough to read transcription tool comparisons with more precision. Accuracy still depends on audio quality and delivery, and proper nouns are frequently misheard — for meeting notes, plan on a human reading the result. It also sits in the history of audio AI as the foundation later work built on.
The easiest way to try it is through a transcription app or web service that embeds it. Developers can also run it locally.