To judge a free Arabic transcription tool, feed it a short clip and read the result against the audio. With Potok the published word error rate for Arabic is 16, so use two clips: one formal, one dialect. Dialect speech scores worse, and the figure comes from other people's tests.
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 16 is a word error rate from published tests of the earlier large-v2 model, not a measurement we made.
- A third-party turbo average of 10.57 exists too, so the published figures for Arabic do not settle on one value.
- Because dialects differ a lot from Modern Standard Arabic, one clip cannot stand for the whole language.
- Arabic runs right to left, which matters when you open the exported file.
What a transcript looks like
[00:00] Speaker 1: لنبدأ الاجتماع. أولاً سنراجع نتائج الأسبوع الماضي، ثم نخطط للخطوات التالية.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 |
|---|---|---|---|
| Arabic (العربية) | Use with care | word error rate 16% (large-v2, FLEURS) Whisper paper, TheWhisper benchmark | Published tests show a higher error rate for this language. Check a short sample first and plan to proofread. The figure is from the earlier large-v2 model and a third-party turbo average was 10.57. Dialects differ a lot from Modern Standard Arabic, and error rates are higher on dialect speech. |
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
- Cut two short clips. Take about half a minute of formal speech, such as a lecture, and the same length of casual talk in your own dialect.
- Open the bot. Open @potok_now_bot in Telegram and send the first clip as a voice message, an audio file or a video.
- Send the second clip. Keep the two results apart so you can compare how the tool copes with each register.
- Read along with the audio. Mark every wrong, missing or invented word. The ratio of marks to words heard is your own rough error rate.
- Scale up only if it holds. For a long recording, files up to 2 GB are accepted, and the Upload button of the bot's mini app handles big ones.
| 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 |
|---|---|---|
| Testing with news-style speech only | The clip looks clean, then family calls come out with many more errors. | Always add one clip of the dialect you actually need. |
| Judging the tool by one lucky sentence | A single easy sentence says little about a full recording. | Use a continuous stretch of speech with a few names in it. |
| Opening the result in a left-to-right editor | Punctuation and digits jump to the wrong side of the line. | Check it in a word processor that supports right-to-left text. |
| Comparing tools with different exports | Differences in layout get mistaken for differences in recognition. | Export the same format from each tool before comparing. |
More
See the full list of languages, the tools and the glossary.
Frequently asked questions
Why does the Arabic page carry a caution?
Published error rates for Arabic are higher than for the recommended languages, and they rise on dialect speech, so a careful read is part of the job.
How long should the test clip be?
Half a minute to a minute is enough to see the pattern of mistakes without spending much time on checking.
Do the published numbers tell me how my audio will do?
No. They come from other people's tests on their own data. Your microphone, noise and dialect decide the real result.
Will vowel marks appear in the text?
Do not count on it. Ordinary Arabic text often leaves them out, so add any marks you need by hand.
Can the text come back in English?
No. The transcript stays in the language of the recording.
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
- Whisper paper, Table 13 (large-v2 on FLEURS) · checked
- TheWhisper benchmark (turbo, average over FLEURS, MLS, Common Voice) · 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