Tajweed is a set of essential rules for Quranic recitation. Ensuring correct pronunciation and efforts in preventing the errors that could alter its meaning is an obligation that needs to be fulfilled. While traditional face-to-face learning is highly effective, it is time-consuming and may not be accessible to all learners. To address this, QuranVision, a real-time Tajweed detection mobile application, is proposed, leveraging Convolutional Neural Networks (CNNs) for image-based Tajweed rule recognition. The study evaluates YOLOv11 and SSD MobileNet V2 FPNLite, focusing on detecting Meem Sakinah-based Tajweed rules (Idgham and Izhar Shafawee). Findings show that YOLOv11 outperforms SSD MobileNet V2 FPNLite, achieving a mean Average Precision (mAP) of 0.723 at IoU 0.5:0.95 with an input size of 416. Prototype testing confirms its effectiveness in detecting Tajweed rules in both black-and-white and color-coded Quranic texts, making it the preferred model for mobile deployment. The application allows users of all backgrounds and ages to self-learn Tajweed through real-time visual and audio feedback, eliminating reliance on continuous instructor supervision. Future enhancements include expanding the dataset, improving model robustness under varied lighting conditions, and integrating speech-based Tajweed correction to further enhance learning experiences.