The transcript looks wrong

It stops after five minutes

That is the Free plan working as designed: it transcribes the first five minutes of each file. The rest of your audio is stored, not lost — upgrading transcribes it without a re-upload. See plans and prices.

It is empty, or nearly

Play the file yourself. Recordings made with the wrong input selected, or with the microphone muted, look like normal files and contain nothing but silence. There is no transcript to produce from silence.

It is in the wrong language

Auto-detect decides from the opening of the file, so an English introduction before an hour of another language sends the whole thing the wrong way. Set the language explicitly and run it again — languages and translation explains when auto-detect is worth overriding.

It is full of mistakes

Listen to thirty seconds of the audio with fresh ears. Distance, echo and people talking over each other explain most of it, and none of them can be fixed after recording. How to get a cleaner transcript covers what actually helps next time; for this file, editing the text directly is the fastest route.

The speakers are wrong

Speaker separation is a best guess and it struggles with overlapping or similar voices. Rename the ones it got right — the name replaces every occurrence — and correct the rest by hand. Speaker labels has the detail.