EdTech · Mobile + Web
tartil.online
An app that listens to the user recite verse by verse and gives instant feedback on exactly which word was off — using a CTC aligner + Whisper trained specifically for the Quran.
The founder, a Quran teacher in Indramayu, had dozens of teenage students who wanted to practice on their own at home — but there was no quick way for them to know which words they were getting wrong. Private lessons are expensive and inflexible; YouTube videos are one-way — no feedback.
General-purpose speech-to-text (Whisper, Google Speech) isn't accurate for classical Arabic recited with tartil — vowel length, articulation points, tajweed and the reciter's style all affect the result. We had to custom-train a Quran-specific alignment pipeline: a CTC aligner that can point at the word level, not the sentence level.
A Flutter frontend records the user's recitation and sends the audio via API to a Laravel backend, which forwards it to a Python pipeline built on sherpa-onnx (a CTC aligner fork) + a fine-tuned Whisper. The output: an array of per-word timestamps + confidence scores. The UI highlights doubtful words, and the user can replay the reference qari and their own recitation side by side.
From brief to TestFlight beta in 6 weeks. 500+ active users in the first 3 months, coming from the founder's WhatsApp study groups. Alignment accuracy of 92% for slow tartil recitation (the main target user). Ongoing: expanding into memorization and murottal evaluation.
Want to see your idea become something like tartil.online?
Start now
30 minutes, free, no commitment. If we click, we move on to a validation session. If not, we'll recommend someone who's a better fit — still at no cost.