Fact · Dictation
The Whisper paper says its evaluation audio covers recording conditions including telephone and VoIP calls and meetings.
Checked against its source on October 16, 2026
In the source's own words
“The audio selection covers various recording conditions such as scripted and unscripted broadcast, telephone and VoIP calls, and meetings.”
Robust Speech Recognition via Large-Scale Weak Supervision, Alec Radford et al., arXiv (eess.AS), 2022, Evaluation data description.
Open the original- What it measures
- The sentence describes the audio used in the paper's evaluation.
- What it does not show
- The paper's test sets are not dictation recordings, and the paper does not report dictation accuracy.