Tarteel.ai needed to give Muslims who are learning Quran recitation real-time, high-accuracy Arabic speech feedback even though many users do not speak Arabic and do not have easy access to qualified instructors.
Reported impact
Tarteel delivered what it describes as the world's first Quranic Arabic ASR system, achieving a 4 percent word-error rate while giving thousands of users live recitation feedback that improves learning and accessibility.
Tarteel built its Quran-recitation assistant on NVIDIA speech AI components instead of a generic transcription stack. Audio recitations from the application are processed by an Arabic automatic speech recognition pipeline fine-tuned with NVIDIA NeMo and served through NVIDIA Riva so users can receive correction while they recite. The team trains and optimizes the models with the NVIDIA CUDA Toolkit on NVIDIA A100 and V100 Tensor Core GPU systems, giving it the GPU-accelerated compute needed to improve accuracy and keep latency low as usage grows. This architecture keeps the model specialized for Quranic Arabic, supports live feedback instead of delayed batch review, and lets Tarteel scale the service across individual learners, instructors, schools, content creators, and researchers.
Users submit Quran recitations through Tarteel, the audio is transcribed by Arabic ASR models built with NeMo, Riva serves the low-latency inference path for live feedback, and the corrected text and recitation guidance are returned in real time inside the application.
"By fine-tuning NVIDIA® Riva automatic speech recognition (ASR) for the Arabic language, we've achieved state-of-the-art WER of 4 percent for Arabic transcription and are able to provide live feedback to recitations, enabling thousands of users to easily learn the Quran."
Provide real-time recitation feedback and correction for Muslims studying the Quran
Tarteel used NVIDIA speech AI tools, including NVIDIA Riva, the GPU-accelerated SDK for building customized speech applications, and NVIDIA NeMo, the open-source framework that enables developers to build and train state-of-the-art conversational AI models.
The platform achieves state-of-the-art results with a 4 percent word-error rate (WER).
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
NVIDIA Sample Oct 6, 2026 University of Pennsylvania Penn needed shared AI infrastructure that could support researchers across many disciplines without weeks-long wait times, fragmented storage, or the cost barriers that kept frontier-scale training and inference off… 1.5 years of compute time reduced to three days for a 40,000-document project NVIDIA DGX B200 systems · NVIDIA DGX SuperPOD · NVIDIA InfiniBand NVIDIA Sample Aug 27, 2026 Texas A&M University Texas A&M needed to give researchers across seven institutions access to shared GPU infrastructure for AI and simulation, but isolated workstations, slow grant-funded clusters, and long queues were limiting… $1M saved on a single project DDN high-performance storage · NVIDIA Base Command Manager · NVIDIA DGX SuperPOD NVIDIA Sample Aug 3, 2026 StudyFetch StudyFetch needed to keep AI tutoring affordable while scaling live lecture transcription and personalized learning for millions of students, but off-the-shelf cloud transcription created cost and latency ceilings that… 100M+ learning interactions NVIDIA H100 GPUs · NVIDIA L40S GPUs · NVIDIA NIM 