Speech-to-text
Also called STT, dictation, transcription, speech recognition
Turning spoken words into written text with an AI model, for dictating messages, transcribing recordings or captioning calls.
Speech recognition used to need you to speak slowly and train it on your voice. That changed when models were trained on huge amounts of real audio. OpenAI's Whisper, released free for anyone to use, learned from 680,000 hours of speech in many languages, accents and noisy rooms.
Now it's built into almost everything: phone keyboards, chat apps, video captions, voicemail and meeting tools. For many people, talking is faster than typing, and newer tools go a step further by removing the "ums" and fixing the grammar as they write it down.
That tidying is useful and a little risky. The cleaner the text, the easier it is to forget it's a machine's reading of what you said.
For example
You talk through your idea for 10 minutes on a walk, and the transcript becomes the first draft of your email.
Put it to use
Where you'll meet speech-to-text in ChatGPT, Gemini and Claude, how to use it there, and the mix-up most people make.
Daily Bytes Plus is in the Android app.