Every word of the recitation, timed to the audio.
Upload a recitation and get back subtitles timed to the word — which surah, which ayah, which word, and exactly when it was said. The text comes from the mushaf, not from the speech recogniser, so what you read is what is written.
- وَرَتِّلِ0:00.42
- ٱلْقُرْءَانَ0:01.24
- تَرْتِيلًا0:02.46
What you get, and what it costs
Free, and there is no account to make. 20 min of audio a day straight away, or 2 h 30 min a day once you have confirmed an email — spend it on one long recitation or a dozen short ones. Files up to 500 MB.
The email is only there because it is the one thing that cannot be reset by opening a private window. We keep a one-way hash of it, never the address itself, and nothing is ever sent to you except the code.
Your recording and its subtitles are deleted 12 hours after the job finishes. Nothing is kept, nothing is used to train anything, and you can delete it sooner from the result page.
The work is paid for by donations, which is why there is a daily limit at all. If you find this useful, a donation raises it for everyone. What it pays for
How it works
Find the place, not the words.Speech recognition is used only to answer where in the Qur'an are we? — never to produce the text you read.
Align the mushaf against the audio.Once the place is known, the canonical text is matched to the recording word by word, so every timestamp comes from the alignment rather than from a guess.
Read it, correct it, take it away.Follow the recitation word by word in the browser, nudge any boundary the alignment got wrong, and download JSON, SRT, VTT, ASS with karaoke timing, per-word SRT, CSV and a Praat TextGrid.
Repeated ayat, a Basmala that may or may not be there, the opening letters no recogniser can read, and speech either side of the recitation are all handled.
Questions people ask
Is it really free?
Yes. There is no account, no card and no trial. A daily allowance keeps the shared bill affordable, and donations are what raise it.
Where does the Arabic text come from?
From the published mushaf, reproduced word for word. Speech recognition is used only to work out which part of the Qur'an a recording is, never to write the text you read — so the words are always exactly what is written.
What can I upload?
Audio or video: mp3, m4a, wav, ogg, mp4, mkv, webm and most other formats, or a YouTube link. Anything with a clear recitation in it will do.
What do I get back?
Seven files: JSON, SRT, VTT, ASS with karaoke timing, a per-word SRT, CSV, and a Praat TextGrid. You get all of them every time, so you never have to know in advance which one you need.
How accurate are the timings?
Usually close enough to use as they are, and occasionally out by a fraction of a second on a long held vowel. That is why the result page lets you play any word and nudge where it starts or ends, and rewrites every file from your correction.
What happens to my recording?
It is deleted a few hours after the job finishes, along with every file made from it. Nothing is kept, nothing is used to train anything, and you can delete it sooner with one button.
Can I use the subtitles in a video I publish?
Yes, including commercially. The only condition travels with the Qur'an text itself, which is used under Creative Commons Attribution 3.0 — and the attribution is already inside every file.
Does it handle repeated ayat?
Yes. Repetition is normal in recitation and in teaching, so a repeated ayah is timed at each occurrence rather than merged, and the output follows the audio in order.