Thai speech to text in Potok carries a caution. The published character error rate is 11.5 for the large-v2 model, and one third-party turbo run gave about 13.4. These are other people's tests, not our measurements. Expect to proofread the result, and send a short clip first.
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
- Thai is written without spaces between words, so quality is reported as a character error rate rather than a word error rate.
- OpenAI itself names Thai as a language where the faster turbo variant gets worse.
- Tone marks are part of the writing, and a wrong or missing mark can produce a different word.
- Without spaces, word breaks in the transcript are the tool's guess, so line breaks may not match how you would split the text.
What a transcript looks like
[00:00] Speaker 1: เรามาเริ่มประชุมกันเลย ก่อนอื่นเราจะทบทวนผลลัพธ์ของสัปดาห์ที่แล้ว จากนั้นจะวางแผนขั้นตอนต่อไปMeaning 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 |
|---|---|---|---|
| Thai (ไทย) | Use with care | character error rate 11.5% (large-v2, FLEURS) Whisper paper, Whale paper, Whisper turbo discussion | Published tests show a higher error rate for this language. Check a short sample first and plan to proofread. Thai is written without spaces, so quality is reported as a character error rate. OpenAI itself names Thai as a language where the faster turbo variant gets worse, and a third-party turbo run gave about 13.4. |
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
- Test a short clip. Send one minute from the real speaker before a long file; with a caution language this saves time.
- Open the bot. Go to @potok_now_bot and send the file, a forwarded voice message or a link.
- Send the long file. Up to 2 GB is accepted; use Upload in the mini app for the largest.
- Read for meaning, not for spaces. Judge a line by whether it says what the speaker said, since spacing is not a reliable guide.
- Check tone marks and numbers. Look at words where one mark changes the meaning, and at any numbers.
- Export. TXT, MD, DOCX, PDF, SRT or VTT.
| 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 |
|---|---|---|
| Comparing counts of words | There are no spaces to count. | Judge by characters or by reading. |
| Choosing the turbo variant elsewhere because it is quicker | Published tests show it gets worse on Thai. | Stay with a larger model for this language. |
| Publishing a first draft | The score is clearly weaker than in recommended languages. | Plan a full proofread. |
| Recording in an echoing room | Echo adds wrong characters. | Use a quiet room. |
More
See the full list of languages, the tools and the glossary.
Frequently asked questions
Is Thai well supported?
It is marked with a caution. The published character error rate is 11.5 for large-v2, so expect to correct the text.
Why character error rate?
Thai has no spaces between words, so counting characters is fairer than counting words.
What about turbo?
OpenAI names Thai as a language where turbo gets worse, and a third party reported about 13.4.
Whose figures are these?
They are published by others. We did not measure Thai ourselves.
Does the bot change the language of the text?
No. The transcript stays in Thai, in the language of the recording, and no other version is made.
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
- Whisper paper, Table 13 (large-v2 on FLEURS) · checked
- Whale paper, Table 5 (Whisper large-v3 on FLEURS) · checked
- OpenAI Whisper discussion on the turbo model · 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