Potok converts Mandarin Chinese audio to text through a Telegram bot. Published tests of others on large-v3, not our measurements, give a character error rate of 7.8, down from 14.7 for the earlier large-v2 model. Chinese has no spaces, so errors are counted per character. 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
- Character error rate counts wrong characters against a correct reference text, so lower is better.
- The earlier large-v2 model scored 14.7 in the same kind of published test.
- Chinese is written without spaces, so the measure counts characters, not words.
- Cantonese is a separate spoken language and has its own page on this site.
What a transcript looks like
[00:00] Speaker 1: 我们开始开会吧。首先回顾上周的结果，然后规划下一步。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 |
|---|---|---|---|
| Chinese (Mandarin) (中文（普通话）) | Recommended | character error rate 7.8% (large-v3, FLEURS) Whale paper, Whisper paper | Published tests show a low error rate for this language. Chinese is written without spaces, so quality is reported as a character error rate. The earlier large-v2 model scored 14.7. |
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
- Confirm the language is Mandarin. If speakers switch to Cantonese or a strong regional dialect, results will drop. Split the file where the language changes.
- Open the bot. Open @potok_now_bot in Telegram.
- Send the file. Audio, video, a voice note or a YouTube link all work, up to 2 GB. Use the Upload button for large ones.
- Proofread sound-alike characters. Different characters can share a pronunciation. Read the result with the audio and fix names in particular.
- Export. Choose DOCX or TXT for editing and SRT for subtitles.
| 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 |
|---|---|---|
| Mixing Mandarin and Cantonese in one file | The recogniser follows one language and the other half degrades. | Cut the file where the language changes. |
| Counting words to measure quality | Text without spaces does not split into words. | Compare characters, as the published figure does. |
| Trusting names | A name can be written with a different but similar-sounding character. | Check names against a written source. |
| Quoting 14.7 as the current result | That number belongs to the earlier large-v2 model. | Use the large-v3 figure of 7.8 and name the model. |
More
See the full list of languages, the tools and the glossary.
Frequently asked questions
How good is Mandarin transcription?
The published character error rate is 7.8 on large-v3, which is in the good group, but names and rare words still need checking.
Why is it a character error rate?
Chinese is written without spaces, so the common measure counts characters rather than words.
Is the large-v2 figure still relevant?
It was 14.7, so the newer model reads better in those published tests.
Can I upload a video?
Yes. Video files and YouTube links are accepted within the 2 GB limit.
Does it write traditional or simplified characters?
The published tests do not say, so check a sample for the form you need.
Sources
Each source was opened by hand on the date shown. Pages change, so open the source if a detail matters to you.
- Word error rate (Wikipedia), Wikipedia · checked
- Whisper speech recognition system (Wikipedia), Wikipedia · checked
- Whale paper, Table 5 (Whisper large-v3 on FLEURS) · checked
- Whisper paper, Table 13 (large-v2 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