Speech recognition

It splits the sound coming in from a microphone into small pieces, works out which speech sounds they are, and writes them out as the most fitting words by looking at the flow before and after. An AI that learned from a huge number of voices paired with their written transcripts does this work. In English it is also called STT (Speech-to-Text).

It is used for phone dictation, video captions, meeting notes, voice assistants and call center records. In a quiet place, when you speak clearly, it is very accurate. But it still makes mistakes in noisy places, when several people talk at once, and with dialects, technical terms and people's names.

It is like having a stenographer who writes down what is said at a meeting inside your phone. It is fast, but it can write the wrong thing when it hears an unfamiliar name or a murmur of voices.

Giving AI commands by voice and automatically turning meetings and calls into text have become common. You should also look at the fact that the transcript can be wrong and at where the recording is stored.