milanakdj/nepali-audio-reserve-r2

Nepali synthetic educational dialogue ~137.0 h of Nepali speech at 24 kHz. Two speakers per clip, 2-5 minutes, teacher/student turns. A backup, not a release: the transcripts are machine-generated, and none of this audio passed the quality gate that produced our training corpus. Synthetic. Generated by a TTS model reading NCERT-style educational dialogue, code-mixed Nepali/English. No human speaker is recorded here. Columns column contents id first 16… See the full description on the dataset page: https://huggingface.co/datasets/milanakdj/nepali-audio-reserve-r2.