Split your Quran recitation, verse by verse
A free, private, and intelligent tool designed for your Quran memorization review (Murajaa). Upload your audio, let the AI detect the verses, adjust, and export — everything happens in your browser.
What is this tool?
Murajaa Surah Splitter lets you take a full surah audio recording and split it into individual verse files (MP3). This is particularly useful for Quran memorization review: you can listen to specific verses on repeat, identify verses you're struggling with, and build a personalized revision playlist.
Whether you recorded yourself reciting, or you have a recitation from a sheikh, this tool will cut it precisely at each verse boundary.
How to use it
Upload your audio
Drag and drop an MP3 file of a full surah recitation. The file stays on your device — nothing is uploaded to any server.
Select the surah
Choose which surah the audio corresponds to. The tool supports both Hafs and Warsh riwayat and knows the exact number of verses for each surah.
Detect or mark the verses
Auto-detect (recommended): The AI model transcribes the Arabic audio and matches it to the Quranic text to find each verse boundary automatically. The first time, it downloads a 131 MB model that is then cached for future use.
Manual mode: Listen to the audio and press Space each time a verse ends. The tool shows the current verse text with Tajweed colors so you know exactly when to cut.
Adjust the markers
After detection or manual marking, fine-tune each verse boundary with the +/- buttons. You can listen to each verse individually to verify the cut is clean. Adjust the step precision from 0.01s to any value you need.
Export
Download a ZIP file containing one MP3 per verse, named sequentially (001.mp3, 002.mp3...). You can optionally prepend the Bismillah with 2 seconds of silence before the first verse.
Key features
100% private
Everything runs locally in your browser. Your audio is never sent to any server. The AI model runs via WebAssembly on your device.
AI-powered verse detection
Uses NVIDIA's FastConformer CTC model, quantized for the browser. It transcribes Arabic speech and uses Levenshtein matching to align the transcription with the Quranic text.
Save your progress
Working on a long surah? Save your session as a draft and come back later. Your audio and markers are stored locally in IndexedDB.
Pre-trim
If your audio has an intro, outro, or unrelated content, set a start and end point before splitting. Only the selected zone will be processed.
Playback speed
Adjust playback speed from 0.25x to 3x. Useful for quickly reviewing long surahs or slowing down to catch precise verse endings.
Tajweed display
Verse text is displayed with Tajweed color coding during marking, so you can follow along with the recitation and know exactly where each verse ends.
Under the hood
The audio is resampled to 16 kHz, then processed in 30-second chunks through a mel spectrogram and the FastConformer ONNX model running on ONNX Runtime Web (WASM). The CTC decoder produces timestamped words, which are matched to the Quranic text using normalized Levenshtein distance.
Audio export uses FFmpeg compiled to WebAssembly (ffmpeg.wasm), allowing precise MP3 splitting without any server-side processing.
The AI model used is open source.GitHub Repo