In a 2024 study of research transcripts, transcribing from scratch and correcting automatic output took equally long, about 7 times real time on average, for the data and conventions tested.

Checked against its source on October 17, 2026

In the source's own words

“transcribing and correcting still take equally long with 7 times real-time on average”

Evaluating Workflows for Creating Orthographic Transcripts for Oral Corpora by Transcribing from Scratch or Correcting ASR Output, Jan Gorisch and Thomas Schmidt, Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), 2024, Abstract.

What it measures: The authors compared manual transcription with correction of automatic output for several kinds of recording. Real time means one hour of audio takes about seven hours of work at that rate.
What it does not show: Research transcripts with strict conventions, on hard recordings such as dialect and historic speech. The authors also say the more complex the data, the more time corrections take. Clean business audio with a good recognizer can behave differently, so the result is a warning, not a rule.

Fact · Transcription

7times real time

Checked against its source on October 17, 2026

“transcribing and correcting still take equally long with 7 times real-time on average”

What it measures: The authors compared manual transcription with correction of automatic output for several kinds of recording. Real time means one hour of audio takes about seven hours of work at that rate.
What it does not show: Research transcripts with strict conventions, on hard recordings such as dialect and historic speech. The authors also say the more complex the data, the more time corrections take. Clean business audio with a good recognizer can behave differently, so the result is a warning, not a rule.