Files
caveman/misc.py
2026-01-10 20:35:21 +01:00

12 lines
309 B
Python

import numpy as np
import librosa
from settings import N_FFT, HOP
def audio_to_logmag(audio):
# STFT
stft = librosa.stft(audio, n_fft=N_FFT, hop_length=HOP)
mag = np.abs(stft)
logmag = np.log1p(mag) # log(1 + x) for stability
return logmag # shape: (freq_bins, time_frames) = (513, T)