@mayfer Maybe because it possessed you
i learn an input spectrogram (magnitude and phase), and parametrise the magnitude so it prefers speech-like audio. It's an inverse 2D-DCT of learnable coefficients scaled by a decay envelope over two axes: quefrency and modulation rate
Users praise the AI researcher's spectrogram optimization for generating speech from listening models because they describe the work as absolutely incredible.
Based on 1 visible X reactions from 1 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@matthen2 Incredible work. Absolutely incredible
i learn an input spectrogram (magnitude and phase), and parametrise the magnitude so it prefers speech-like audio. It's an inverse 2D-DCT of learnable coefficients scaled by a decay envelope over two axes: quefrency and modulation rate
@mayfer Maybe because it possessed you
Users praise the AI researcher's spectrogram optimization for generating speech from listening models because they describe the work as absolutely incredible.
Based on 1 visible X reactions from 1 accounts; directional sample.
Ask a question below.
Published answers will appear here.
