Skip to main content

Answer · Hard audio

How do I transcribe audio with strong accents?

Choose the spoken language yourself instead of relying on auto-detect, add names, places and jargon to the names and terms box, and upload the clearest recording you have. Most accents transcribe well on clear audio; errors cluster on names, numbers and words that sound like a more common word. Check those against the audio before quoting.

Try it on your own file

Free for 3 files a day, up to 30 minutes each. Speaker labels, timestamps, and exports for Word, SRT and NVivo.

Drop audio or video here

MP3, MP4, M4A, WAV, MOV, WEBM, OGG · up to 5 hours

What actually goes wrong

Speech recognition models learn from large amounts of speech, and some accents are better represented than others. With an accent the model hears less often, it tends to swap a rare word for a common word that sounds close.

The classic example is a technical term turned into an unrelated, sometimes embarrassing, everyday word. The sentence then reads smoothly but says the wrong thing, which is why you should check key terms rather than skim.

Mixed languages

Many speakers switch languages mid-sentence, such as Hindi and English, or Spanish and English. Pick the language most of the recording is in. Words in the other language usually still come out, but check them.

If one long section is entirely in another language, splitting that section into its own file and choosing its language gives better results.

The details

Pick the language

For short clips or recordings that start with small talk, auto-detection can guess wrong. Choosing the language removes that risk.

Terms box for names

Names of people, places, drugs or products are where accents cause the most errors. Type them in before upload.

Clear audio matters more with accents

An accent plus noise is much harder than either alone. Record close to the speaker.

Ask speakers to spell key names

In an interview, ask the person to spell their name and organisation once. You get the spelling for the terms box and a reference in the audio.

Check the smooth-sounding sentences

A misheard word often produces a grammatical sentence that is wrong. Read for meaning, not only for typos.

Common mistakes

  • Leaving the language on auto for a 30-second clip.
  • Translating the audio in your head while transcribing: keep the transcript in the language spoken, then use translation if you need it.
  • Assuming an accurate first page means the whole file is accurate. Accuracy changes with each speaker.

Questions people also ask

Does it work with Indian, Nigerian or Scottish English?

Yes, English with these and other accents is transcribed. Accuracy depends more on audio clarity and how many rare terms there are than on the accent itself. Try a free file to see.

Can I transcribe Hinglish or Spanglish?

Choose the main language of the recording. Words from the other language usually still appear, sometimes misspelled. Check them, and add common ones to the terms box.

Can I translate the transcript afterwards?

Yes, on a paid plan you can translate a finished transcript into other languages. Transcribe in the spoken language first for the best result.

Which languages are supported?

About 40, including English, Spanish, Portuguese, French, German, Hindi, Arabic, Chinese, Japanese, Korean, Turkish, Vietnamese and Indonesian.

Related questions