Fact · Languages and accents
117,000hours
Of Whisper's 680,000 hours of training audio, 117,000 hours cover 96 languages other than English.
Checked against its source on October 17, 2026
In the source's own words
“Of those 680,000 hours of audio, 117,000 hours cover 96 other languages.”
Robust Speech Recognition via Large-Scale Weak Supervision (full text), Alec Radford et al., OpenAI (arXiv), 2022, Introduction.
Open the original (DOI 10.48550/arXiv.2212.04356)- What it measures
- Hours per language are very uneven, which the paper's later analysis relates to accuracy.
- What it does not show
- Hours of audio in the training set are not hours of verified transcript. The authors collected the data from the web, and some transcripts are noisy.