Cantonese transcription: audio and video to text

By the Potok teamUpdated ~2 min read

Cantonese is a borderline language for Potok: it works through the Telegram bot, but treat the text as a draft. Published tests of the large-v3 model report a character error rate of 9.8, and a vendor test of the turbo variant reported much worse results. Files up to 2 GB are accepted.

Transcribe a recording

Send your voice message, audio or video to @potok_now_bot on Telegram and the text comes back in the chat.

Open @potok_now_bot in Telegram

Files up to 2 GB · export to TXT, MD, DOCX, PDF, SRT and VTT · speaker labels and timestamps · about 30 languages

Key facts

  • Cantonese is written without spaces, so quality is reported as a character error rate: 9.8 in published tests of large-v3.
  • A vendor test reported much worse results for the turbo variant, which is why we mark Cantonese as borderline.
  • Character error rate and word error rate are different scales, so do not compare 9.8 with a figure for English.
  • Cantonese and Mandarin are separate entries on our language list.

What a transcript looks like

Sample sentence in Cantonese
[00:00] Speaker 1: 我哋開始開會啦。首先睇下上個禮拜嘅結果,跟住傾下一步點做。

Meaning in English: Let's start the meeting. First, we'll review last week's results, then plan the next steps.

Written by us to show the script and the layout of a transcript. It is not the output of the recognition engine.

Published quality

Quality for this language
LanguageVerdictPublished resultNotes
Cantonese (粵語)Recommended, borderlinecharacter error rate 9.8% (large-v3, FLEURS)
Whale paper
Published tests show an error rate close to our 15% line. Check a short sample first.
Cantonese is written without spaces, so quality is reported as a character error rate. A vendor test reported much worse results for the faster turbo variant, so treat this language as borderline.

These numbers come from published tests by other people, not from tests by Potok. Your own audio, with its noise, accents and microphone, will differ.

How to do it

  1. Try a short clip. Because the language is borderline, send a minute of your own speaker to @potok_now_bot before the full file.
  2. Judge the clip. Read the Cantonese text beside the audio. Look at colloquial words, place names and any English words mixed into the speech.
  3. Send the real recording. If the draft is usable, send the whole file. Up to 2 GB works; the Upload button in the mini app handles the largest ones.
  4. Add timestamps for review. With timestamps on, you can jump back to a doubtful line and replay it.
  5. Export for subtitles or notes. SRT and VTT give timed lines for video; TXT, MD, DOCX and PDF suit a document.
What Potok can do
FeatureDetails
Where it worksTelegram bot @potok_now_bot; large files go through the bot's mini app
File sizeUp to 2 GB
ExportTXT, MD, DOCX, PDF, SRT, VTT
Speakers and timeSpeaker labels and timestamps can be switched on
LanguagesAbout 30 languages with good published results; a few more with warnings
TranslationThe transcript is not translated into another language
DictationA separate Potok desktop app lets you dictate text into any program; it recognises Russian speech

Common mistakes

Common mistakes and how to fix them
MistakeWhat happensFix
Using a Mandarin expectation for a Cantonese recordingColloquial vocabulary is easy to get wrong and you may blame the wrong cause.Check the result with someone who reads written Cantonese.
Comparing the 9.8 character rate with a word rateThe two numbers use different units, so the comparison misleads.Compare only figures of the same metric.
Trusting the transcript for subtitles without reviewWrong characters stay on screen for viewers.Proofread the SRT before you publish the video.
Sending a noisy street recordingA borderline language loses more under noise than a strong one.Use the cleanest audio you can get.

More

See the full list of languages, the tools and the glossary.

Frequently asked questions

Why is Cantonese marked as borderline?

The character error rate of 9.8 looks fine, yet a vendor test reported much worse results for the turbo variant, so we advise a careful check.

What is a character error rate?

It is the share of characters that differ from a reference text. It is used for languages written without spaces, where counting words is not practical.

Does the bot separate spoken Cantonese from Mandarin?

They are separate entries on our list, but the published tests give no detail on mixed speech, so test a clip if your audio mixes them.

Can I get subtitles for a Cantonese video?

Yes. Export SRT or VTT, then proofread the lines before you publish them.

Is the transcript translated into English?

No. The text stays in Cantonese because the bot does not translate transcripts.

Sources

Each source was opened by hand on the date shown. Pages change, so open the source if a detail matters to you.

  1. Word error rate (Wikipedia), Wikipedia · checked
  2. Whisper speech recognition system (Wikipedia), Wikipedia · checked
  3. Whale paper, Table 5 (Whisper large-v3 on FLEURS) · checked

Try it on your own recording

Send your voice message, audio or video to @potok_now_bot on Telegram and the text comes back in the chat.

Open @potok_now_bot in Telegram