Qira'a Studio model assets
This repository provides the two inference archives used by Qira'a Studio. The app pins a revision and checks the archives and their contents before loading them.
studio-models.zip: additional acoustic classifiers, CNNs, BiLSTM, CRNN, and Bayesian model parameters.studio-xlsr.zip: the five fine-tuned XLS-R checkpoint deltas used by the app's ensemble.
These are deployment assets for a research demonstration, not a new performance benchmark. The models offer tentative suggestions for the ten Quranic readings and do not verify correctness or identify reciters.
This repository does not contain the exact RBF SVM, its support vectors, training recordings, feature tables, or per-clip predictions.