dlyog/af_heart_arm_tts_dataset

af_heart_arm_tts_dataset The distillation corpus used to train dlyog/af_heart_arm_tts — a single-voice TTS model built for Arm inference on NVIDIA DGX Spark. 3.0000 hours · 4,104 clips · 24 kHz mono 16-bit PCM, synthesized by Kokoro-82M speaking af_heart. This is knowledge distillation: the teacher generated every clip, so the student can approach it but never exceed it. The pipeline that produced this — and that reproduces it for any other voice — is at… See the full description on the dataset page: https://huggingface.co/datasets/dlyog/af_heart_arm_tts_dataset.