First comprehensive benchmark for evaluating Nepali text-to-speech systems.
sentences.json — 365 phonologically-designed Nepali sentences with metadata
audio/ — TTS outputs from 12 systems (~2000 audio files)
results/ — Pre-computed evaluation results (SCOREQ, Chirp2, MMS, XLS-R, Whisper)
baselines.json — Aggregate scores for all baseline systems
model/ — NepaliMOS predictor checkpoint (Spearman 0.587)
Systems Evaluated… See the full description on the dataset page: https://huggingface.co/datasets/ampixa/neptts-bench. 
