Dutch speech to text with Potok is a recommended language. Published tests of large-v3 give a word error rate of 5.57, one of the lower numbers on our list. Some users report that large-v3 does worse on their own Dutch data than the older model, so check a sample.
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
- 5.57 is a published figure from others, not something we measured on your files.
- Reports from users, not studies, suggest that large-v3 can lose to the earlier model on some Dutch material.
- Dutch is spoken in the Netherlands and in Belgium, and accents differ between the two.
- A low score on clean test speech does not carry over to noisy rooms.
- Both the Netherlands and Belgium use the language, so accent is a fair thing to test.
What a transcript looks like
[00:00] Speaker 1: Laten we de vergadering beginnen. Eerst bespreken we de resultaten van vorige week, daarna plannen we de volgende stappen.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 |
|---|---|---|---|
| Dutch (Nederlands) | Recommended | word error rate 5.57% (large-v3, FLEURS) Canary-1B-v2 paper | Published tests show a low error rate for this language. Some users report that large-v3 does worse than the earlier model on their own Dutch data. |
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
- Pick your hardest minute. Choose a stretch with overlap or noise rather than a calm opening. It tells you more.
- Send it. Open @potok_now_bot in Telegram and attach it.
- Compare to the source. Count the wrong words and note which kind they are: names, numbers or run-together words.
- Decide. If you are happy, send the full file, up to 2 GB, through Upload if it is big.
- Export. Pick 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 |
|---|---|---|
| Trusting the low number blindly | The user reports mentioned above show it will not hold for every file. | Read a sample first. |
| Testing only the easy intro | You miss the failure mode. | Choose the noisy section. |
| Ignoring the accent of the speaker | Belgian and Dutch speech can differ in word choice. | Test with the real speaker. |
| Treating compound words as errors | Dutch writes long words joined; a check for gaps may flag correct text. | Compare against the audio. |
More
See the full list of languages, the tools and the glossary.
Frequently asked questions
What is the published Dutch score?
A word error rate of 5.57 for the large-v3 model, from published tests.
What is the catch?
Some users report that large-v3 does worse than the earlier model on their own Dutch data.
Is the 5.57 yours?
No. It comes from tests published by others.
Does Flemish work?
The figure does not break out accents, so try a short clip from the speaker.
Can I get a PDF?
Yes. PDF is one of the six export formats, along with TXT, MD, DOCX, SRT and VTT.
Can I test the older-model claim myself?
Not through the bot, but you can judge the output by reading a hard clip against the audio and counting errors.
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
- FLEURS speech dataset (Google, Hugging Face), Hugging Face · checked
- Canary-1B-v2 paper, Table 11 (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