mirror of
https://github.com/Anjok07/ultimatevocalremovergui.git
synced 2024-11-24 15:30:11 +01:00
253 lines
9.9 KiB
Batchfile
253 lines
9.9 KiB
Batchfile
@echo off
|
|
|
|
cls
|
|
:start
|
|
ECHO =======================================
|
|
ECHO Multi-Model Processer and Ensembler
|
|
ECHO =======================================
|
|
ECHO.
|
|
ECHO 1. Ensemble 4Band Model Outputs - 44100Hz
|
|
ECHO 2. Ensemble 3Band Model Outputs - 44100Hz
|
|
ECHO 3. Ensemble 2Band Model Outputs - 32000Hz
|
|
ECHO 4. Exit
|
|
set choice=
|
|
set /p choice=Type the number associated with your choice and hit 'Enter':
|
|
if not '%choice%'=='' set choice=%choice:~0,1%
|
|
if '%choice%'=='1' goto model1
|
|
if '%choice%'=='2' goto model2
|
|
if '%choice%'=='3' goto model3
|
|
if '%choice%'=='4' goto end
|
|
if '%choice%'=='' goto end
|
|
ECHO "%choice%" is not valid, try again
|
|
ECHO.
|
|
goto start
|
|
:model1
|
|
ECHO MODEL MGM-v5-4Band-44100-BETA1
|
|
set model=MGM-v5-4Band-44100-BETA1
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\4band_44100.json -w 352 -P models\%model%.pth -t -i %1
|
|
|
|
ECHO MODEL MGM-v5-4Band-44100-BETA2
|
|
set model=MGM-v5-4Band-44100-BETA2
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\4band_44100.json -w 352 -P models\%model%.pth -t -i %1
|
|
|
|
ECHO MODEL HighPrecison_4band_1
|
|
set model=HighPrecison_4band_1
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\4band_44100.json -w 352 -n 123821KB -P models\%model%.pth -t -i %1
|
|
|
|
ECHO MODEL HighPrecison_4band_2
|
|
set model=HighPrecison_4band_2
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\4band_44100.json -w 352 -n 123821KB -P models\%model%.pth -t -i %1
|
|
|
|
ECHO MODEL BigLayer_4band_1
|
|
set model=BigLayer_4band_1
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\4band_44100.json -w 352 -n 129605KB -P models\%model%.pth -t -i %1
|
|
|
|
ECHO MODEL BigLayer_4band_2
|
|
set model=BigLayer_4band_2
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\4band_44100.json -w 352 -n 129605KB -P models\%model%.pth -t -i %1
|
|
|
|
ECHO MODEL BigLayer_4band_3
|
|
set model=BigLayer_4band_3
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\4band_44100.json -w 352 -n 129605KB -P models\%model%.pth -t -i %1
|
|
|
|
ECHO Ensembling Instruments...
|
|
set modelparam=1band_sr44100_hl512
|
|
cd /d %~dp0
|
|
|
|
python lib/spec_utils.py -a min_mag -m modelparams\%modelparam%.json separated\%~n1_MGM-v5-4Band-44100-BETA1_Instruments.wav separated\%~n1_MGM-v5-4Band-44100-BETA2_Instruments.wav -o ensembled/temp/"1E2E_ensam1"
|
|
|
|
python lib/spec_utils.py -a min_mag -m modelparams\%modelparam%.json separated\%~n1_HighPrecison_4band_1_Instruments.wav separated\%~n1_HighPrecison_4band_2_Instruments.wav -o ensembled/temp/"3E4E_ensam2"
|
|
|
|
python lib/spec_utils.py -a min_mag -m modelparams\%modelparam%.json separated\%~n1_BigLayer_4band_1_Instruments.wav separated\%~n1_BigLayer_4band_2_Instruments.wav -o ensembled/temp/"5E6E_ensam3"
|
|
|
|
python lib/spec_utils.py -a min_mag -m modelparams\%modelparam%.json ensembled/temp/"1E2E_ensam1"_v.wav ensembled/temp/"3E4E_ensam2"_v.wav -o ensembled/temp/"1E2E3E4E_ensam4"
|
|
|
|
python lib/spec_utils.py -a min_mag -m modelparams\%modelparam%.json ensembled/temp/"5E6E_ensam3"_v.wav ensembled/temp/"1E2E3E4E_ensam4"_v.wav -o ensembled/temp/"A6_ensam5"
|
|
|
|
python lib/spec_utils.py -a min_mag -m modelparams\%modelparam%.json separated\%~n1_BigLayer_4band_3_Instruments.wav ensembled/temp/"A6_ensam5"_v.wav -o ensembled/"%~n1_4BAND_Ensembled_Instrumental"
|
|
|
|
del ensembled\temp\"1E2E_ensam1"_v.wav
|
|
del ensembled\temp\"3E4E_ensam2"_v.wav
|
|
del ensembled\temp\"5E6E_ensam3"_v.wav
|
|
del ensembled\temp\"1E2E3E4E_ensam4"_v.wav
|
|
del ensembled\temp\"A6_ensam5"_v.wav
|
|
|
|
ECHO Ensembling Vocals...
|
|
|
|
set modelparam=1band_sr44100_hl512
|
|
cd /d %~dp0
|
|
|
|
python lib/spec_utils.py -a max_mag -m modelparams\%modelparam%.json separated\%~n1_MGM-v5-4Band-44100-BETA1_Vocals.wav separated\%~n1_MGM-v5-4Band-44100-BETA2_Vocals.wav -o ensembled/temp/"1EV2EV_ensam1"
|
|
|
|
python lib/spec_utils.py -a max_mag -m modelparams\%modelparam%.json separated\%~n1_HighPrecison_4band_1_Vocals.wav separated\%~n1_HighPrecison_4band_2_Vocals.wav -o ensembled/temp/"3EV4EV_ensam2"
|
|
|
|
python lib/spec_utils.py -a max_mag -m modelparams\%modelparam%.json separated\%~n1_BigLayer_4band_1_Vocals.wav separated\%~n1_BigLayer_4band_2_Vocals.wav -o ensembled/temp/"5EV6EV_ensam3"
|
|
|
|
python lib/spec_utils.py -a max_mag -m modelparams\%modelparam%.json ensembled/temp/"1EV2EV_ensam1"_v.wav ensembled/temp/"3EV4EV_ensam2"_v.wav -o ensembled/temp/"1EV2EV3EV4EV_ensam4"
|
|
|
|
python lib/spec_utils.py -a max_mag -m modelparams\%modelparam%.json ensembled/temp/"5EV6EV_ensam3"_v.wav ensembled/temp/"1EV2EV3EV4EV_ensam4"_v.wav -o ensembled/temp/"A6V_ensam5"
|
|
|
|
python lib/spec_utils.py -a max_mag -m modelparams\%modelparam%.json separated\%~n1_BigLayer_4band_3_Vocals.wav ensembled/temp/"A6V_ensam5"_v.wav -o ensembled/"%~n1_4BAND_Ensembled_Vocals"
|
|
|
|
del ensembled\temp\"1EV2EV_ensam1"_v.wav
|
|
del ensembled\temp\"3EV4EV_ensam2"_v.wav
|
|
del ensembled\temp\"5EV6EV_ensam3"_v.wav
|
|
del ensembled\temp\"1EV2EV3EV4EV_ensam4"_v.wav
|
|
del ensembled\temp\"A6V_ensam5"_v.wav
|
|
|
|
set choice=
|
|
set /p choice=Delete Individual Model Outputs? [Y/N]:
|
|
if not '%choice%'=='' set choice=%choice:~0,1%
|
|
if '%choice%'=='Y' goto delete
|
|
if '%choice%'=='y' goto delete
|
|
if '%choice%'=='N' goto keep
|
|
if '%choice%'=='n' goto keep
|
|
if '%choice%'=='' goto keep
|
|
ECHO "%choice%" is not valid, try again
|
|
ECHO.
|
|
goto start
|
|
:delete
|
|
del separated\%~n1_MGM-v5-4Band-44100-BETA1_Instruments.wav
|
|
del separated\%~n1_MGM-v5-4Band-44100-BETA2_Instruments.wav
|
|
del separated\%~n1_HighPrecison_4band_1_Instruments.wav
|
|
del separated\%~n1_HighPrecison_4band_2_Instruments.wav
|
|
del separated\%~n1_BigLayer_4band_1_Instruments.wav
|
|
del separated\%~n1_BigLayer_4band_2_Instruments.wav
|
|
del separated\%~n1_BigLayer_4band_3_Instruments.wav
|
|
del separated\%~n1_MGM-v5-4Band-44100-BETA1_Vocals.wav
|
|
del separated\%~n1_MGM-v5-4Band-44100-BETA2_Vocals.wav
|
|
del separated\%~n1_HighPrecison_4band_1_Vocals.wav
|
|
del separated\%~n1_HighPrecison_4band_2_Vocals.wav
|
|
del separated\%~n1_BigLayer_4band_1_Vocals.wav
|
|
del separated\%~n1_BigLayer_4band_2_Vocals.wav
|
|
del separated\%~n1_BigLayer_4band_3_Vocals.wav
|
|
ECHO Complete!
|
|
goto start
|
|
:keep
|
|
ECHO Complete!
|
|
goto end
|
|
:model2
|
|
ECHO MODEL MGM-v5-MIDSIDE-44100-BETA1
|
|
set model=MGM-v5-MIDSIDE-44100-BETA1
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\3band_44100.json -w 352 -P models\%model%.pth -t -i %1
|
|
|
|
ECHO MODEL MGM-v5-MIDSIDE-44100-BETA2
|
|
set model=MGM-v5-MIDSIDE-44100-BETA2
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\3band_44100.json -w 352 -P models\%model%.pth -t -i %1
|
|
|
|
ECHO MODEL MGM-v5-3Band-44100-BETA
|
|
set model=MGM-v5-3Band-44100-BETA
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\3band_44100.json -w 352 -P models\%model%.pth -t -i %1
|
|
|
|
ECHO Ensembling Instruments...
|
|
set modelparam=1band_sr44100_hl512
|
|
cd /d %~dp0
|
|
|
|
python lib/spec_utils.py -a min_mag -m modelparams\%modelparam%.json separated\%~n1_MGM-v5-MIDSIDE-44100-BETA1_Instruments.wav separated\%~n1_MGM-v5-MIDSIDE-44100-BETA2_Instruments.wav -o ensembled/temp/"1E2E_ensam1"
|
|
|
|
python lib/spec_utils.py -a min_mag -m modelparams\%modelparam%.json separated\%~n1_MGM-v5-3Band-44100-BETA_Instruments.wav ensembled/temp/"1E2E_ensam1"_v.wav -o ensembled/"%~n1_3BAND_Ensembled_Instrumental"
|
|
|
|
del ensembled\temp\"1E2E_ensam1"_v.wav
|
|
|
|
ECHO Ensembling Vocals...
|
|
|
|
set modelparam=1band_sr44100_hl512
|
|
cd /d %~dp0
|
|
|
|
python lib/spec_utils.py -a max_mag -m modelparams\%modelparam%.json separated\%~n1_MGM-v5-MIDSIDE-44100-BETA1_Vocals.wav separated\%~n1_MGM-v5-MIDSIDE-44100-BETA2_Vocals.wav -o ensembled/temp/"1EV2EV_ensam1"
|
|
|
|
python lib/spec_utils.py -a max_mag -m modelparams\%modelparam%.json separated\%~n1_MGM-v5-3Band-44100-BETA_Vocals.wav ensembled/temp/"1EV2EV_ensam1"_v.wav -o ensembled/"%~n1_3BAND_Ensembled_Vocals"
|
|
|
|
del ensembled\temp\"1EV2EV_ensam1"_v.wav
|
|
|
|
set choice=
|
|
set /p choice=Delete Individual Model Outputs? [Y/N]:
|
|
if not '%choice%'=='' set choice=%choice:~0,1%
|
|
if '%choice%'=='Y' goto delete
|
|
if '%choice%'=='y' goto delete
|
|
if '%choice%'=='N' goto keep
|
|
if '%choice%'=='n' goto keep
|
|
if '%choice%'=='' goto keep
|
|
ECHO "%choice%" is not valid, try again
|
|
ECHO.
|
|
goto start
|
|
:delete
|
|
del separated\%~n1_MGM-v5-MIDSIDE-44100-BETA1_Instruments.wav
|
|
del separated\%~n1_MGM-v5-MIDSIDE-44100-BETA2_Instruments.wav
|
|
del separated\%~n1_MGM-v5-3Band-44100-BETA_Instruments.wav
|
|
del separated\%~n1_MGM-v5-MIDSIDE-44100-BETA1_Vocals.wav
|
|
del separated\%~n1_MGM-v5-MIDSIDE-44100-BETA2_Vocals.wav
|
|
del separated\%~n1_MGM-v5-3Band-44100-BETA_Vocals.wav
|
|
ECHO Complete!
|
|
goto start
|
|
:keep
|
|
ECHO Complete!
|
|
goto end
|
|
:model3
|
|
ECHO MODEL MGM-v5-2Band-32000-BETA1
|
|
set model=MGM-v5-2Band-32000-BETA1
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\2band_32000.json -w 352 -P models\%model%.pth -t -i %1
|
|
|
|
ECHO MODEL MGM-v5-2Band-32000-BETA2
|
|
set model=MGM-v5-2Band-32000-BETA2
|
|
cd /d %~dp0
|
|
|
|
python inference.py -g 0 -m modelparams\2band_32000.json -w 352 -P models\%model%.pth -t -i %1
|
|
|
|
ECHO Ensembling Instruments...
|
|
set modelparam=1band_sr44100_hl512
|
|
cd /d %~dp0
|
|
|
|
python lib/spec_utils.py -a min_mag -m modelparams\%modelparam%.json separated\%~n1_MGM-v5-2Band-32000-BETA1_Instruments.wav separated\%~n1_MGM-v5-2Band-32000-BETA2_Instruments.wav -o ensembled/"%~n1_2BAND_Ensembled_Instrumental"
|
|
|
|
ECHO Ensembling Vocals...
|
|
|
|
set modelparam=1band_sr44100_hl512
|
|
cd /d %~dp0
|
|
|
|
python lib/spec_utils.py -a max_mag -m modelparams\%modelparam%.json separated\%~n1_MGM-v5-2Band-32000-BETA1_Vocals.wav separated\%~n1_MGM-v5-2Band-32000-BETA2_Vocals.wav -o ensembled/"%~n1_2BAND_Ensembled_Vocals"
|
|
|
|
set choice=
|
|
set /p choice=Delete Individual Model Outputs? [Y/N]:
|
|
if not '%choice%'=='' set choice=%choice:~0,1%
|
|
if '%choice%'=='Y' goto delete
|
|
if '%choice%'=='y' goto delete
|
|
if '%choice%'=='N' goto keep
|
|
if '%choice%'=='n' goto keep
|
|
if '%choice%'=='' goto keep
|
|
ECHO "%choice%" is not valid, try again
|
|
ECHO.
|
|
goto start
|
|
:delete
|
|
del separated\%~n1_MGM-v5-2Band-32000-BETA1_Instruments.wav
|
|
del separated\%~n1_MGM-v5-2Band-32000-BETA2_Instruments.wav
|
|
del separated\%~n1_MGM-v5-2Band-32000-BETA1_Vocals.wav
|
|
del separated\%~n1_MGM-v5-2Band-32000-BETA2_Vocals.wav
|
|
ECHO Complete!
|
|
goto start
|
|
:keep
|
|
ECHO Complete!
|
|
goto end
|
|
:end |