Instructions to use Muno459/fastconformer-quran with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- NeMo
How to use Muno459/fastconformer-quran with NeMo:
import nemo.collections.asr as nemo_asr asr_model = nemo_asr.models.ASRModel.from_pretrained("Muno459/fastconformer-quran") transcriptions = asr_model.transcribe(["file.wav"]) - Notebooks
- Google Colab
- Kaggle
License question — non-profit foundation use (employee Quran program, Indonesia)
Assalamu'alaikum, and thank you for this model — and for the honesty note about the EveryAyah leakage. Correcting your own headline number downward is why I trust the rest of them.
I'd like to ask whether our use fits NPL-1.0 before we touch your weights.
Who we are. Yayasan Darul Wahyu, a registered Islamic foundation in Indonesia. We can provide our legal deed (akta yayasan) if that's useful.
What we're building. A Quran reading app for a workplace recitation program. It uses forced alignment (CTC posteriors, constrained Viterbi over the known mushaf text) to record which words were actually voiced. It does not grade recitation quality or tajweed — it only verifies that a page was genuinely read aloud.
There is an Indonesian tradition called patungan or arisan mengaji: a group divides the Quran among its members — a few pages or a juz each — so that together they complete a khatam. Today this runs entirely on trust and self-reporting. The app lets a group track it honestly.
Funding and commercial status, stated plainly. The readers are employees of DRW Skincare, and our foundation is funded through that company's CSR budget, which pays for the server and the development. The app itself is never sold: no ads, no paid tier, no revenue of any kind. Nobody pays to use it. We would not be redistributing your weights.
I'm disclosing the corporate funding link rather than describing us only as a foundation, so you can judge it accurately rather than on a flattering half-picture.
Does this fall within NPL-1.0? If it doesn't, we'll respect that and stay on the base NVIDIA model — no hard feelings either way, and thank you for the work regardless.
Wa alaykum as-salam wa rahmatullah. Jazak Allah khayran for one of the clearest and most honest disclosures I have received, especially for naming the CSR funding link yourself instead of just saying "a foundation". That is exactly what lets me answer you properly.
Short answer: yes, what you have described fits NPL-1.0.
What the licence cares about is that no revenue is derived from the model or from the app around it. You have stated plainly: nothing is sold, no ads, no paid tier, nobody pays to use it, and you are not redistributing the weights. Corporate CSR money paying for the server and the development is covering cost, not earning from it, and that is the distinction the licence draws. So you are within the terms.
Two conditions it rests on, just so they are written down:
- it stays free to every end user. If ads, a paid tier, or any revenue is ever added later, it leaves the licence and we should talk first.
- keep attribution to NVIDIA for the upstream base and to the datasets, as you already planned.
No need to send the akta yayasan, your description is enough.
And thank you for the kind words about the leakage correction. A number that flatters me but misleads you is worth nothing.
The patungan / arisan mengaji use is a beautiful thing to build for. May Allah accept it from your foundation and from your readers.