Fact · Transcription
20 to 100milliseconds tolerance (stricter test)
Under a stricter tolerance of 20 to 100 milliseconds, the 2025 study's word alignments from Whisper were more accurate than those of earlier work.
Checked against its source on October 17, 2026
In the source's own words
“produces word alignments that are more accurate than prior work under a stricter tolerance between 20 ms and 100 ms”
Whisper Has an Internal Word Aligner, Sung-Lin Yeh et al., arXiv (ASRU 2025), 2025, Abstract.
Open the original- What it measures
- Tighter tolerances show whether a timestamp is useful for captions or audio-linked editing, where a few hundred milliseconds matter.
- What it does not show
- Compares against prior work under the stricter setting, on the authors' test data. Not a universal accuracy figure.