Answer
Hashchats transcribes the existing audio rather than regenerating speech. This recovery adds no AI usage charge. A loading indicator appears while timing is prepared; if it fails, retry while ordinary audio remains playable.
Recovered text can differ from the original message formatting and may contain transcription errors. Once timing is saved, participants can reuse it. Recovery is requested per narration; older audio is not automatically processed in bulk.
Replaying, seeking and opening the reader after timing is saved make no new AI request. Existing valid timing stays in place, so an older narration may highlight its title and opening together.