Cool_Tools/ultimatevocalremovergui
1
0
Fork 0
mirror of https://github.com/Anjok07/ultimatevocalremovergui.git synced 2024-11-28 01:10:56 +01:00
Code Issues Packages Projects Releases Wiki Activity

Update spec_utils.py

Browse Source
This commit is contained in:
aufr33 2021-07-01 10:38:49 +03:00 committed by GitHub
parent f7cb34201f
commit 0b79e8e991
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 49 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
lib/spec_utils.py
View File

@ -285,6 +285,55 @@ def cmb_spectrogram_to_wave(spec_m, mp, extra_bins_h=None, extra_bins=None):
return wave.T
def cmb_spectrogram_to_wave_ffmpeg(spec_m, mp, tmp_basename, extra_bins_h=None, extra_bins=None):
import subprocess
bands_n = len(mp.param['band'])
offset = 0
ffmprc = {}
for d in range(1, bands_n + 1):
bp = mp.param['band'][d]
spec_s = np.ndarray(shape=(2, bp['n_fft'] // 2 + 1, spec_m.shape[2]), dtype=complex)
h = bp['crop_stop'] - bp['crop_start']
spec_s[:, bp['crop_start']:bp['crop_stop'], :] = spec_m[:, offset:offset+h, :]
tmp_wav = '{}_cmb_spectrogram_to_wave_b{}_sr{}'.format(tmp_basename, d, str(bp['sr']) + '.wav')
tmp_wav2 = '{}_cmb_spectrogram_to_wave_b{}_sr{}'.format(tmp_basename, d, str(mp.param['sr']) + '.wav')
offset += h
if d == bands_n: # high-end
if extra_bins_h: # if --high_end_process bypass
max_bin = bp['n_fft'] // 2
spec_s[:, max_bin-extra_bins_h:max_bin, :] = extra_bins[:, :extra_bins_h, :]
if bp['hpf_start'] > 0:
spec_s = fft_hp_filter(spec_s, bp['hpf_start'], bp['hpf_stop'] - 1)
if bands_n == 1:
wave = spectrogram_to_wave(spec_s, bp['hl'], mp, False)
else:
wave = spectrogram_to_wave(spec_s, bp['hl'], mp, False)
else:
if d == 1: # lower
spec_s = fft_lp_filter(spec_s, bp['lpf_start'], bp['lpf_stop'])
else: # mid
spec_s = fft_hp_filter(spec_s, bp['hpf_start'], bp['hpf_stop'] - 1)
spec_s = fft_lp_filter(spec_s, bp['lpf_start'], bp['lpf_stop'])
sf.write(tmp_wav, spectrogram_to_wave(spec_s, bp['hl'], mp, False).T, bp['sr'])
ffmprc[d] = subprocess.Popen(['ffmpeg', '-hide_banner', '-loglevel', 'panic', '-y', '-i', tmp_wav, '-ar', str(mp.param['sr']), '-ac', '2', '-c:a', 'pcm_s16le', tmp_wav2])
for s in ffmprc:
ffmprc[s].communicate()
for d in range(bands_n - 1, 0, -1):
os.remove(f'{tmp_basename}_cmb_spectrogram_to_wave_b{d}_sr' + str(mp.param['band'][d]['sr']) + '.wav')
tmp_wav2 = f'{tmp_basename}_cmb_spectrogram_to_wave_b{d}_sr' + str(mp.param['sr']) + '.wav'
wave2, _ = librosa.load(tmp_wav2, mp.param['sr'], False, dtype=np.float32, res_type="sinc_fastest")
os.remove(tmp_wav2)
wave = np.add(wave, wave2)
return wave.T
def fft_lp_filter(spec, bin_start, bin_stop):
g = 1.0
for b in range(bin_start, bin_stop):