Spaces:
Running
Running
| import numpy | |
| def normalize(audio): | |
| audio=audio-(numpy.min(audio)+numpy.max(audio))/2 | |
| return audio*(1-(max(numpy.max(audio), abs(numpy.min(audio))))) | |
| def pitch(audio, pitch, grain): | |
| grain=int(grain) | |
| if len(audio)>10: audio=[audio] | |
| if type(audio) is not list: audio=audio.tolist() | |
| length=len(audio[0]) | |
| if pitch<1: | |
| pitch=int(1//pitch) | |
| if grain%2!=0: grain+=1 | |
| for i in range(len(audio)): | |
| n=0 | |
| while n+grain<length: | |
| #print(len(audio[i])) | |
| #print(n) | |
| audio[i][n:n+grain]=numpy.repeat(audio[i][n:n+int(grain/2)], 2) | |
| #print(len(audio[i])) | |
| n+=grain | |
| elif pitch>1: | |
| pitch=int(pitch) | |
| for i in range(len(audio)): | |
| n=0 | |
| while n+grain<length: | |
| audio[i][n:n+grain]=audio[i][n:n+grain:pitch]*pitch | |
| n+=grain | |
| return audio | |
| def pitchB(audio, pitch, grain): | |
| grain=int(grain) | |
| if len(audio)>10: audio=[audio] | |
| if type(audio) is not list: audio=audio.tolist() | |
| length=len(audio[0]) | |
| if pitch<1: | |
| pitch=int(1//pitch) | |
| if grain%2!=0: grain+=1 | |
| for i in range(len(audio)): | |
| n=0 | |
| while n+grain<length: | |
| #print(len(audio[i])) | |
| #print(n) | |
| audio[i][n:n+grain]=numpy.repeat(audio[i][n:n+int(grain/2)], 2) | |
| #print(len(audio[i])) | |
| n+=grain | |
| elif pitch>1: | |
| pitch=int(pitch) | |
| for i in range(len(audio)): | |
| n=0 | |
| while n+grain<length: | |
| audio2=audio[i][n:n+grain:pitch] | |
| for j in range(pitch-1): | |
| #print(j) | |
| audio2.extend(audio2[::1] if j%2==1 else audio2[::-1]) | |
| audio[i][n:n+grain]=audio2 | |
| n+=grain | |
| return audio | |
| def grain(audio, grain): | |
| grain=int(grain) | |
| if len(audio)>10: audio=[audio] | |
| if type(audio) is not list: audio=audio.tolist() | |
| length=len(audio[0]) | |
| n=0 | |
| for i in range(len(audio)): | |
| while n+2*grain<length: | |
| audio[i][n+grain:n+2*grain]=audio[i][n:n+grain] | |
| n+=grain*2 | |
| return audio | |
| def ftt(audio, inverse=True): | |
| """headphone warning: cursed effect""" | |
| import scipy.fft | |
| audio=numpy.asarray(audio).copy() | |
| for i in range(len(audio)): | |
| if inverse is False: | |
| audio[i]= scipy.fft.fft(audio[i], axis=0) | |
| else: | |
| audio[i]= scipy.fft.ifft(audio[i], axis=0) | |
| audio=(audio*(2/numpy.max(audio)))-1 | |
| return normalize(audio) | |
| def fourier_shift(audio, value=5): | |
| """modulates volume for some reason""" | |
| import scipy.ndimage | |
| audio=numpy.asarray(audio).copy() | |
| audio= numpy.asarray(list(scipy.ndimage.fourier_shift(i, value, axis=-1) for i in audio)).astype(float) | |
| return normalize(audio) | |
| def gradient(audio): | |
| """acts as an interesting high pass filter that removes drums""" | |
| audio=numpy.asarray(audio).copy() | |
| return numpy.gradient(audio, axis=0) | |
| def gradient_inverse(audio): | |
| """supposed to be inverse of a gradient, but it just completely destroys the audio into a distorted mess""" | |
| audio=numpy.asarray(audio).copy() | |
| for i in range(len(audio)): | |
| a = audio[i] | |
| audio[i] = a[0] + 2 * numpy.c_[numpy.r_[0, a[1:-1:2].cumsum()], a[::2].cumsum() - a[0] / 2].ravel()[:len(a)] | |
| audio=normalize(audio) | |
| return numpy.gradient(audio, axis=0) | |