heimoshuiyu/ultimatevocalremovergui
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add files via upload

Browse Source
This commit is contained in:
Anjok07
2022-07-03 18:47:33 -05:00
committed by GitHub
parent 7cce5e303a
commit 04f29d3bb7
3 changed files with 489 additions and 77 deletions
Download Patch File Download Diff File Expand all files Collapse all files

157
inference_v5.py
View File

@@ -11,6 +11,12 @@ import numpy as np
import soundfile as sf
from tqdm import tqdm
from demucs.pretrained import get_model as _gm
from demucs.hdemucs import HDemucs
from demucs.apply import BagOfModels, apply_model
from pathlib import Path
from models import stft, istft
from lib_v5 import dataset
from lib_v5 import spec_utils
from lib_v5.model_param_init import ModelParameters
@@ -51,7 +57,13 @@ data = {
'window_size': 512,
'agg': 10,
'high_end_process': 'mirroring',
'ModelParams': 'Auto'
'ModelParams': 'Auto',
'demucsmodel_sel_VR': 'UVR_Demucs_Model_1',
'overlap': 0.5,
'shifts': 0,
'segment': 'None',
'split_mode': False,
'demucsmodelVR': True,
}
default_window_size = data['window_size']
@@ -97,6 +109,11 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
global nn_arch_sizes
global nn_architecture
global overlap_set
global shift_set
global split_mode
global demucs_model_set
#Error Handling
runtimeerr = "CUDNN error executing cudnnSetTensorNdDescriptor"
@@ -140,8 +157,14 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
# For instrumental the instrumental is the temp file
# and for vocal the instrumental is the temp file due
# to reversement
if data['demucsmodelVR']:
sameplerate = 44100
else:
sameplerate = mp.param['sr']
sf.write(f'temp.wav',
wav_instrument, mp.param['sr'])
wav_instrument.T, sameplerate)
appendModelFolderName = modelFolderName.replace('/', '_')
@@ -176,14 +199,14 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
if VModel in model_name and data['voc_only']:
sf.write(instrumental_path,
wav_instrument, mp.param['sr'])
wav_instrument.T, sameplerate)
elif VModel in model_name and data['inst_only']:
pass
elif data['voc_only']:
pass
else:
sf.write(instrumental_path,
wav_instrument, mp.param['sr'])
wav_instrument.T, sameplerate)
# Vocal
if vocal_name is not None:
@@ -215,14 +238,14 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
if VModel in model_name and data['inst_only']:
sf.write(vocal_path,
wav_vocals, mp.param['sr'])
wav_vocals.T, sameplerate)
elif VModel in model_name and data['voc_only']:
pass
elif data['inst_only']:
pass
else:
sf.write(vocal_path,
wav_vocals, mp.param['sr'])
wav_vocals.T, sameplerate)
if data['saveFormat'] == 'Mp3':
try:
@@ -362,6 +385,11 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
text_widget.clear()
button_widget.configure(state=tk.DISABLED) # Disable Button
overlap_set = float(data['overlap'])
shift_set = int(data['shifts'])
demucs_model_set = data['demucsmodel_sel_VR']
split_mode = data['split_mode']
vocal_remover = VocalRemover(data, text_widget)
modelFolderName = determineModelFolderName()
@@ -369,6 +397,7 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
try: #Load File(s)
for file_num, music_file in enumerate(data['input_paths'], start=1):
# Determine File Name
m=music_file
base_name = f'{data["export_path"]}/{file_num}_{os.path.splitext(os.path.basename(music_file))[0]}'
model_name = os.path.basename(data[f'{data["useModel"]}Model'])
@@ -802,6 +831,85 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
y_spec_m = pred * X_phase
v_spec_m = X_spec_m - y_spec_m
def demix_demucs(mix):
#print('shift_set ', shift_set)
text_widget.write(base_text + "Running Demucs Inference...\n")
text_widget.write(base_text + "Processing... ")
print(' Running Demucs Inference...')
mix = torch.tensor(mix, dtype=torch.float32)
ref = mix.mean(0)
mix = (mix - ref.mean()) / ref.std()
with torch.no_grad():
sources = apply_model(demucs, mix[None], split=split_mode, device=device, overlap=overlap_set, shifts=shift_set, progress=False)[0]
text_widget.write('Done!\n')
sources = (sources * ref.std() + ref.mean()).cpu().numpy()
sources[[0,1]] = sources[[1,0]]
return sources
def demucs_prediction(m):
global demucs_sources
mix, samplerate = librosa.load(m, mono=False, sr=44100)
if mix.ndim == 1:
mix = np.asfortranarray([mix,mix])
mix = mix.T
demucs_sources = demix_demucs(mix.T)
if data['demucsmodelVR']:
demucs = HDemucs(sources=["other", "vocals"])
text_widget.write(base_text + 'Loading Demucs model... ')
update_progress(**progress_kwargs,
step=0.95)
path_d = Path('models/Demucs_Models')
print('What Demucs model was chosen? ', demucs_model_set)
demucs = _gm(name=demucs_model_set, repo=path_d)
text_widget.write('Done!\n')
print('segment: ', data['segment'])
if data['segment'] == 'None':
segment = None
if isinstance(demucs, BagOfModels):
if segment is not None:
for sub in demucs.models:
sub.segment = segment
else:
if segment is not None:
sub.segment = segment
else:
try:
segment = int(data['segment'])
if isinstance(demucs, BagOfModels):
if segment is not None:
for sub in demucs.models:
sub.segment = segment
else:
if segment is not None:
sub.segment = segment
text_widget.write(base_text + "Segments set to "f"{segment}.\n")
except:
segment = None
if isinstance(demucs, BagOfModels):
if segment is not None:
for sub in demucs.models:
sub.segment = segment
else:
if segment is not None:
sub.segment = segment
print('segment port-process: ', segment)
demucs.cpu()
demucs.eval()
demucs_prediction(m)
if data['voc_only'] and not data['inst_only']:
pass
else:
@@ -809,13 +917,25 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
if data['high_end_process'].startswith('mirroring'):
input_high_end_ = spec_utils.mirroring(data['high_end_process'], y_spec_m, input_high_end, mp)
wav_instrument = spec_utils.cmb_spectrogram_to_wave(y_spec_m, mp, input_high_end_h, input_high_end_)
if data['demucsmodelVR']:
wav_instrument = spec_utils.cmb_spectrogram_to_wave_d(y_spec_m, mp, input_high_end_h, input_high_end_, demucs=True)
demucs_inst = demucs_sources[0]
sources = [wav_instrument,demucs_inst]
spec = [stft(sources[0],2048,1024),stft(sources[1],2048,1024)]
ln = min([spec[0].shape[2], spec[1].shape[2]])
spec[0] = spec[0][:,:,:ln]
spec[1] = spec[1][:,:,:ln]
v_spec_c = np.where(np.abs(spec[1]) <= np.abs(spec[0]), spec[1], spec[0])
wav_instrument = istft(v_spec_c,1024)
else:
wav_instrument = spec_utils.cmb_spectrogram_to_wave_d(y_spec_m, mp, input_high_end_h, input_high_end_, demucs=False)
if data['voc_only'] and not data['inst_only']:
pass
else:
text_widget.write('Done!\n')
else:
wav_instrument = spec_utils.cmb_spectrogram_to_wave(y_spec_m, mp)
wav_instrument = spec_utils.cmb_spectrogram_to_wave_d(y_spec_m, mp)
if data['voc_only'] and not data['inst_only']:
pass
else:
@@ -828,14 +948,25 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
if data['high_end_process'].startswith('mirroring'):
input_high_end_ = spec_utils.mirroring(data['high_end_process'], v_spec_m, input_high_end, mp)
wav_vocals = spec_utils.cmb_spectrogram_to_wave(v_spec_m, mp, input_high_end_h, input_high_end_)
if data['demucsmodelVR']:
wav_vocals = spec_utils.cmb_spectrogram_to_wave_d(v_spec_m, mp, input_high_end_h, input_high_end_, demucs=True)
demucs_voc = demucs_sources[1]
sources = [wav_vocals,demucs_voc]
spec = [stft(sources[0],2048,1024),stft(sources[1],2048,1024)]
ln = min([spec[0].shape[2], spec[1].shape[2]])
spec[0] = spec[0][:,:,:ln]
spec[1] = spec[1][:,:,:ln]
v_spec_c = np.where(np.abs(spec[1]) >= np.abs(spec[0]), spec[1], spec[0])
wav_vocals = istft(v_spec_c,1024)
else:
wav_vocals = spec_utils.cmb_spectrogram_to_wave_d(v_spec_m, mp, input_high_end_h, input_high_end_, demucs=False)
if data['inst_only'] and not data['voc_only']:
pass
else:
text_widget.write('Done!\n')
else:
wav_vocals = spec_utils.cmb_spectrogram_to_wave(v_spec_m, mp)
else:
wav_vocals = spec_utils.cmb_spectrogram_to_wave_d(v_spec_m, mp, demucs=False)
if data['inst_only'] and not data['voc_only']:
pass
else:
@@ -843,7 +974,7 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
update_progress(**progress_kwargs,
step=1)
# Save output music files
save_files(wav_instrument, wav_vocals)