It is much easier to correct a transcription than to generate it wholesale. As well, the task of audio transcription correction has long since been commoditized because of the deployment of speech recognition on every smartphone.

It’s not quite a solved problem but it’s close.