WebAudio input

WebAudio FFT

Model input, 2 second log-scale Mel spectrogram

Predicted language scores (top -)