Fact · Languages and accents
OpenAI says Whisper's performance varies widely depending on the language, and it publishes a per-language breakdown of error rates.
Checked against its source on October 16, 2026
In the source's own words
“Whisper's performance varies widely depending on the language.”
openai/whisper: Robust Speech Recognition via Large-Scale Weak Supervision, OpenAI, GitHub (OpenAI repository), 2023, Available models and languages section.
Open the original- What it measures
- The README points to a per-language chart of word error rates and character error rates, measured on the Common Voice 15 and Fleurs datasets.
- What it does not show
- The sentence says the variation is wide. It does not give a number for Bengali, Urdu or Arabic, so check the chart for the model you plan to use.
Source
Cited on
- Dictation Languages: What Each App and Device SupportsGuide
- What Is Whisper AI? OpenAI's Speech Model ExplainedGuide
- Hindi Voice Typing and Speech to TextGuide
- How to Run Whisper Locally on Your Own ComputerGuide
- Speech Recognition in Python: Libraries ComparedGuide
- Transcribe and Translate Audio in Other LanguagesGuide
- Can you voice type in Bangla, Urdu or Arabic on Windows?Question
- Is speech recognition less accurate in languages other than English?Question
- What languages does Whisper support?Question