To compare free Cantonese transcription tools, run one short clip through each and count the wrong characters. Potok relies on a published character error rate of 9.8 for Cantonese with large-v3, but this language is borderline, so treat that number as other people's result.
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
- The 9.8 is a character error rate (CER) from a published test of large-v3, not our own measurement.
- Cantonese is written without spaces, so quality is counted in characters rather than words.
- A vendor test reported much worse results for the faster turbo variant, so Cantonese stays borderline.
- The transcript stays in Cantonese and is not translated.
What a transcript looks like
[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
| Language | Verdict | Published result | Notes |
|---|---|---|---|
| Cantonese (粵語) | Recommended, borderline | character 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
- Choose a short clip. Use about half a minute of clear Cantonese speech with one speaker, so mistakes are easy to trace.
- Open the bot. Open @potok_now_bot in Telegram and send the clip as a voice message, an audio file or a video.
- Count characters, not words. Without spaces there are no word boundaries. Copy the text into a sheet and mark each wrong or missing character.
- Compare tools on the same clip. Send the identical file to every tool you try. Differences then come from the tool, not from the audio.
- Move to a longer file. Files up to 2 GB are accepted, and the Upload button of the bot's mini app handles the large ones. Export TXT, DOCX or SRT.
| Feature | Details |
|---|---|
| Where it works | Telegram bot @potok_now_bot; large files go through the bot's mini app |
| File size | Up to 2 GB |
| Export | TXT, MD, DOCX, PDF, SRT, VTT |
| Speakers and time | Speaker labels and timestamps can be switched on |
| Languages | About 30 languages with good published results; a few more with warnings |
| Translation | The transcript is not translated into another language |
| Dictation | A separate Potok desktop app lets you dictate text into any program; it recognises Russian speech |
Common mistakes
| Mistake | What happens | Fix |
|---|---|---|
| Counting words in a Cantonese result | There are no spaces, so the count is meaningless. | Count characters and divide by the total heard. |
| Trusting one number for borderline results | Cantonese sits near the edge of what we call reliable. | Treat every transcript as a draft and re-read it. |
| Mixing speakers in the test clip | Overlap makes it hard to see which errors belong to which voice. | Test one voice first, then add a second. |
| Skipping a second tool | With one result there is nothing to compare against. | Run the same clip elsewhere and keep both texts side by side. |
More
See the full list of languages, the tools and the glossary.
Frequently asked questions
Why is Cantonese marked borderline?
The published figure is decent, yet a vendor test reported much worse results for the turbo variant, so we advise caution.
What is a character error rate?
It counts wrong, missing or extra characters against a human reference. It suits languages written without spaces.
Is 9.8 what I will get?
Not necessarily. It comes from a published test on other audio. Your speaker, noise and vocabulary change the outcome.
Can I use timestamps to check long passages?
Yes. Switch timestamps on, then jump to the spot in the audio where a doubtful phrase sits.
Is the output converted to Mandarin?
No. The transcript stays in the language of the recording and is not changed into another one.
Sources
Each source was opened by hand on the date shown. Pages change, so open the source if a detail matters to you.
- Whisper speech recognition system (Wikipedia), Wikipedia · checked
- Word error rate (Wikipedia), Wikipedia · checked
- 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