patyupin/whispernotes-coreml

WhisperNotes CoreML models

On-device (Apple Neural Engine / GPU) CoreML conversions used by WhisperNotes.

prithiv_gender_fp16.mlpackage.zip

Voice-gender recommender — a CoreML fp16 conversion of prithivMLmods/Common-Voice-Gender-Detection (Apache-2.0), converted with coremltools.

• Input input_values: 3 s / 48000-sample mono 16 kHz waveform, wav2vec2 zero-mean / unit-variance normalized.
• Output: 2-class softmax, id2label = {0: female, 1: male}.
• Used only as a soft gender recommendation for new speakers (abstain < 0.75); never a hard matching gate.

Redistributed under Apache-2.0 with attribution to the original author.