jU>SSKrSSKrSSKrSSKrSSKrSSKJr Srg)N)metadatac [R"[5n[R"5n[R"SSS9nUR U5 UR U5 [R"SSS9nURSSS [RS 9 [R"S 5RnS nS nSnSnSn URS5n U RSSSSU3US9 U RSSSUS9 U RSSSUS9 U RSSSUS9 U RSS U S!9 U RS"S#S$9 U RS%[S&S'9 U RS(S)S*/S)S+S,9 S-n S.n S/n S0nS1nS2nS3nURS45nURS5S6S7U S!9 URS8S9S:U S;9 URSS:US!9 URS?S:US!9 URS@SAUS!9 URSBSUS9 SCnSDnSEnSFnSGnSHnSInSJnSKnSLnSMnSNnSOnSPn SQn!SRn"URSS5n#U#RSTSUS9 U#RSU[ SVUSW9 U#RSX[ SYUSW9 U#RSZS:US!9 U#RS[[S\USW9 U#RS]SUS9 U#R#5n$U$RS^SUS9 U$RS_SUS9 U#RS`SUS9 U#RSaSUS9 U#RSb[ S:USW9 U#RScS:/SdQUSe9 U#RSfS9[ S:USg9 U#RShS:U S!9 U#RSiSU!S9 U#RSj[$R&S:U"SW9 Skn%Sln&Smn'Snn(Son)URSp5n*U*RSq[SrU%SW9 U*RSs[ StU&SW9 U*RSu[SvU'SW9 U*RSw[SxU(SW9 U*RSySU)S9 Szn+S{n,S|n-S}n.S~n/Sn0Sn1URS5n2U2RS[SvU+SW9 U2RS[SU,SW9 U2RS[SU-SW9 U2RSSU.S9 U2RSSU/S9 U2RSSU0S9 U2RS[ SU1SW9 Sn3Sn4Sn5Sn6URS5n7U7RS[(SU3SW9 U7RS[SU4SW9 U7RS[ StU5SW9 U7RS[*SU6SW9 Sn8Sn9Sn:Sn;SnU>R.(a[R0n?O([3[U>R4R755n?UR9U?5 U>R:(a#SSKJn@ U@"5nA[@RB"S5 U>RD(Ga9SSKJn@ U@"SS9nAU>RFS*:Xa0WARI5nB[K[$RL"UBSS95 GOWAROU>RPS9nCU>RR(a?U>RRS:a/[U[WWCRY55S:U>RR5nC[[[]S5[[SWCR_5555nD[[[]S5[[SUCRa5555nE[[[]S5[[SUCRa5555nF[[[]S5[[SUCRa5555nGUDUE-UF-UG-S-nH[KSUH-5 [KSSUD3SSSUE3SSSUF3SS35 [KSUH-5 UCRY5HBunInJSRcUJS5nK[KUISWD3SUJSSWE3SUKSWF3SUJS35 MD [@RB"S5 U>Rd(GaSSKJn@ U@"SS9nAUARg5nLUL(d![KS5 [@RB"S5 [[[]S5[[SWLR_5555nM[[[]S5[[SULRa5555nN[]S5nO[[[]S5[[SULRa5555nPUMUN-UO-UP-S-nH[KSUH-5 [KSSUM3SSSUN3SSSUO3SS35 [KSUH-5 ULRY5HjunQnRURRiSS5S:SnS[]URRiS/55nTURRiSS5nU[KUQSWM3SUSSWN3SUTSWO3SUU35 Ml [@RB"S5 U>Rj(aSSKJn@ U>Rl/U>Rn=(d /-nVW@"UU?U>RpS9nAUVH8nWURsSUSUWSU>Rp35 WARuUW5 M: URsSSRcWV5S35 [@RB"S5 [W[3U>S/55nXUX(d&URw5 [@RB"Sv5 URsSUSSRcWX535 SSKJn@ U@"S0SU_SU?_SU>Rp_SU>Rx_SU>Rz_SU>R|_SU>R~_SU>R_SU>R_SU>R_SU>R_SU>R_SU>R_SU>R_SU>R_SU>R_SU>R_SU>R_SU>R_SU>R_SU>RU>RU>RU>RU>RS._SU>RU>RU>RU>RU>RU>RU>RS._SU>RU>RU>RU>RS._SU>RU>RU>RU>RU>RS._6nAU>R(a2U>RlS7:Xa"U>Rn(dWAR5 OFU>Rl/U>Rn=(d /-nY[]UY5Sv:XaWYSnYWARWYS9 WARWXU>RS9nZWZ(d'URS5 [@RB"Sv5 URsSSRcWZ535 g:![a3n[URSU[3SS9 [@RB"Sv5 S:n[A[NS:n[A[ff=f)zMain entry point for the CLI.zB%(asctime)s.%(msecs)03d - %(levelname)s - %(module)s - %(message)sz%Y-%m-%d %H:%M:%S)fmtdatefmtz)Separate audio file into different stems.c,[R"USS9$)N<)max_help_position)argparseRawTextHelpFormatter)progs ڧC:\Users\johnn\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.13_qbz5n2kfra8p0\LocalCache\local-packages\Python313\site-packages\audio_separator/utils/cli.pymain..s,|D|Y|YZ^rt|u) descriptionformatter_class audio_files*zDThe audio file paths or directory to separate, in any common format.)nargshelpdefaultzaudio-separatorz+Show the program's version number and exit.z6Enable debug logging, equivalent to --log_level=debug.z'Print environment information and exit.zzList all supported models and exit. Use --list_filter to filter/sort the list and --list_limit to show only top N results.z). Example: --output_dir=/app/separatedzModel files directory (default: %(default)s or AUDIO_SEPARATOR_MODEL_DIR env var if set). Example: --model_file_dir=/app/modelszADownload a single model file only, without performing separation.zSeparation I/O Paramsz-mz--model_filenamez)model_bs_roformer_ep_317_sdr_12.9755.ckptz--extra_models+N)rrrz--output_formatFLACz--output_bitratez --output_dirz--model_file_dirz/tmp/audio-separator-models/z--download_model_onlyzWInvert secondary stem using spectrogram (default: %(default)s). Example: --invert_spectznMax peak amplitude to normalize input and output audio to (default: %(default)s). Example: --normalization=0.7zlMin peak amplitude to amplify input and output audio to (default: %(default)s). Example: --amplification=0.4zzOutput only single stem, e.g. Instrumental, Vocals, Drums, Bass, Guitar, Piano, Other. Example: --single_stem=Instrumentalz_Modify the sample rate of the output audio (default: %(default)s). Example: --sample_rate=44100zTUse soundfile to write audio output (default: %(default)s). Example: --use_soundfilezSUse PyTorch autocast when supported (default: %(default)s). Example: --use_autocastzUse native float16 for verified model/device combinations (default: %(default)s). Mutually exclusive with --use_autocast. Example: --use_native_fp16zCompile verified repeated model blocks when supported (default: %(default)s). Best for long inputs or repeated same-shape runs; a fresh compiler cache can make the first run slower. Example: --use_torch_compilezUse DirectML for hardware-accelerated inference on Windows AMD/Intel GPUs (experimental; requires the 'dml' extra). Example: --use_directmla'Split audio into chunks of this duration in seconds (default: %(default)s = no chunking). Useful for processing very long audio files on systems with limited memory. Recommended: 600 (10 minutes) for files >1 hour. Chunks are concatenated without overlap/crossfade. Example: --chunk_duration=600zAlgorithm to use for ensembling multiple models (default: avg_wave). Choices: avg_wave, median_wave, min_wave, max_wave, avg_fft, median_fft, min_fft, max_fft, uvr_max_spec, uvr_min_spec, ensemble_wav. Example: --ensemble_algorithm=uvr_max_speczWeights for ensembling multiple models (default: equal). Number of weights must match number of models. Example: --ensemble_weights 1.0 0.5zUse a named ensemble preset (e.g. vocal_balanced, karaoke). Presets define models + algorithm. Use --list_presets to see all. Example: --ensemble_preset=vocal_balancedz-List all available ensemble presets and exit.zCustom names for all output files in JSON format (default: %(default)s). Example: --custom_output_names='{"Vocals": "vocals_output", "Drums": "drums_output"}'zCommon Separation Parametersz--invert_spectz--normalizationg?)rrrz--amplificationgz --single_stemz --sample_rateiDz--use_soundfilez--use_autocastz--use_native_fp16z--use_torch_compilez--use_directmlz--chunk_durationz--ensemble_algorithm) avg_wave median_wavemin_wavemax_waveavg_fft median_fftmin_fftmax_fft uvr_max_spec uvr_min_spec ensemble_wav)rrrz--ensemble_weights)rrrrz--ensemble_presetz--list_presetsz--custom_output_nameszsLarger consumes more resources, but may give better results (default: %(default)s). Example: --mdx_segment_size=256zAmount of overlap between prediction windows, 0.001-0.999. Higher is better but slower (default: %(default)s). Example: --mdx_overlap=0.25zlLarger consumes more RAM but may process slightly faster (default: %(default)s). Example: --mdx_batch_size=4zUsually called stride in neural networks, only change if you know what you're doing (default: %(default)s). Example: --mdx_hop_length=1024zXEnable denoising during separation (default: %(default)s). Example: --mdx_enable_denoisezMDX Architecture Parametersz--mdx_segment_sizez --mdx_overlapg?z--mdx_batch_sizez--mdx_hop_lengthiz--mdx_enable_denoisezNumber of batches to process at a time. Higher = more RAM, slightly faster processing (default: %(default)s). Example: --vr_batch_size=16zBalance quality and speed. 1024 = fast but lower, 320 = slower but better quality. (default: %(default)s). Example: --vr_window_size=320zIntensity of primary stem extraction, -100 - 100. Typically, 5 for vocals & instrumentals (default: %(default)s). Example: --vr_aggression=2ziEnable Test-Time-Augmentation; slow but improves quality (default: %(default)s). Example: --vr_enable_ttazgMirror the missing frequency range of the output (default: %(default)s). Example: --vr_high_end_processzIdentify leftover artifacts within vocal output; may improve separation for some songs (default: %(default)s). Example: --vr_enable_post_processzlThreshold for post_process feature: 0.1-0.3 (default: %(default)s). Example: --vr_post_process_threshold=0.1zVR Architecture Parametersz--vr_batch_sizez--vr_window_sizeiz--vr_aggressionz--vr_enable_ttaz--vr_high_end_processz--vr_enable_post_processz--vr_post_process_thresholdg?zSize of segments into which the audio is split, 1-100. Higher = slower but better quality (default: %(default)s). Example: --demucs_segment_size=256zNumber of predictions with random shifts, higher = slower but better quality (default: %(default)s). Example: --demucs_shifts=4zOverlap between prediction windows, 0.001-0.999. Higher = slower but better quality (default: %(default)s). Example: --demucs_overlap=0.25z_Enable segment-wise processing (default: %(default)s). Example: --demucs_segments_enabled=FalsezDemucs Architecture Parametersz--demucs_segment_sizeDefaultz--demucs_shiftsz--demucs_overlapz--demucs_segments_enabledTztLarger consumes more resources, but may give better results (default: %(default)s). Example: --mdxc_segment_size=256zyOverride model default segment size instead of using the model default value. Example: --mdxc_override_model_segment_sizezNumber of overlapping prediction windows, 2-50. Higher is better but slower (default: model config, falling back to 8). Example: --mdxc_overlap=8zLarger consumes more RAM but may process slightly faster (default: model config, falling back to 1). Example: --mdxc_batch_size=4zShift audio pitch by a number of semitones while processing. May improve output for deep/high vocals. (default: %(default)s). Example: --mdxc_pitch_shift=2zMDXC Architecture Parametersz--mdxc_segment_sizez"--mdxc_override_model_segment_sizez--mdxc_overlapz--mdxc_batch_sizez--mdxc_pitch_shiftr) Separator) info_only)indent)filter_sort_byzModel Filenamec38# UHn[U5v M g7fNlen).0filenames r main..s;hZghCMMZgArchc3># UHn[US5v M g7f)TypeNr8r:rs r r<r=s-\ODc$v,.?.?OzOutput Stems (SDR)c3\# UH"n[SRUS55v M$ g7f), StemsN)r9joinrBs r r<r=s)# UHn[US5v M g7f)NameNr8rBs r r<r=s6eUdTs4<7H7HUdrC- Descriptionc3^# UH#n[URSS5SS5v M% g7f)rNrr9getr:ps r r<r=s-0n]mXYQUU="5Mcr5R1S1S]ms+-Models Algorithmc3X# UH n[URSS55v M" g7f) algorithmrQNrRrTs r r<r=s%.eTdqs155b3I/J/JTds(* rrQrmodelsrY) log_formatter log_levelmodel_file_dirzSeparator version z downloading model z to directory zModel z downloaded successfully.z beginning with input path(s): r\r]r^ output_dir output_formatoutput_bitratenormalization_thresholdamplification_thresholdoutput_single_steminvert_using_spec sample_rate use_soundfile use_autocastuse_native_fp16use_torch_compile use_directmlchunk_durationensemble_algorithmensemble_weightsensemble_preset mdx_params) hop_length segment_sizeoverlap batch_sizeenable_denoise vr_params)rt window_size aggression enable_ttaenable_post_processpost_process_thresholdhigh_end_process demucs_params)rrshiftsrssegments_enabled mdxc_params)rrrtrsoverride_model_segment_size pitch_shift)model_filename)custom_output_nameszSeparation failed: )exc_infou9Separation produced no output files — see errors above.z%Separation complete! Output file(s):  )glogging getLogger__name__ StreamHandler Formatter setFormatter addHandlerr ArgumentParser add_argumentSUPPRESSr distributionradd_argument_groupintfloatadd_mutually_exclusive_grouprloadsstrbool parse_argsdebugDEBUGgetattrr]uppersetLevelenv_infoaudio_separator.separatorr2sysexit list_models list_formatlist_supported_model_filesprintdumpsget_simplified_model_list list_filter list_limitdictlistitemsmaxr9keysvaluesrG list_presetslist_ensemble_presetsrSdownload_model_onlyr extra_modelsr^rdownload_model_and_data print_helpr_r`ra normalization amplification single_stem invert_spectrfrgrhrirjrkrlrmrnromdx_hop_lengthmdx_segment_size mdx_overlapmdx_batch_sizemdx_enable_denoise vr_batch_sizevr_window_size vr_aggression vr_enable_ttavr_enable_post_processvr_post_process_thresholdvr_high_end_processdemucs_segment_size demucs_shiftsdemucs_overlapdemucs_segments_enabledmdxc_segment_sizemdxc_batch_size mdxc_overlap mdxc_override_model_segment_sizemdxc_pitch_shift load_modelseparater Exceptionerror)\logger log_handlerr\parserpackage_version version_help debug_help env_info_helplist_models_helplog_level_help info_paramsmodel_filename_helpextra_models_helpoutput_format_helpoutput_bitrate_helpoutput_dir_helpmodel_file_dir_helpdownload_model_only_help io_paramsinvert_spect_helpnormalization_helpamplification_helpsingle_stem_helpsample_rate_helpuse_soundfile_helpuse_autocast_helpuse_native_fp16_helpuse_torch_compile_helpuse_directml_helpchunk_duration_helpensemble_algorithm_helpensemble_weights_helpensemble_preset_helplist_presets_helpcustom_output_names_help common_paramsprecision_paramsmdx_segment_size_helpmdx_overlap_helpmdx_batch_size_helpmdx_hop_length_helpmdx_enable_denoise_helprpvr_batch_size_helpvr_window_size_helpvr_aggression_helpvr_enable_tta_helpvr_high_end_process_helpvr_enable_post_process_helpvr_post_process_threshold_helprvdemucs_segment_size_helpdemucs_shifts_helpdemucs_overlap_helpdemucs_segments_enabled_helpr}mdxc_segment_size_help%mdxc_override_model_segment_size_helpmdxc_overlap_helpmdxc_batch_size_helpmdxc_pitch_shift_helprargsr]r2 separator model_listr[filename_width arch_width stems_width name_width total_widthr;rstemspresetsid_width desc_width models_width algo_width preset_idpresetdesc num_modelsalgomodels_to_downloadmodelrmodel_filenames output_fileses\ r mainr& s   x (F'')K%%*nyLMM]+ k"  $ $1\ouvF  S7}HPHYHYZ++,=>FFO@LIJ=MTSN++,@AKT;yIVeUfJgnz{T9\ ST< =YT?4oN -7U\op 223DT__EF!%5!6;hZ`ZeZeZg;h8hiNS[#-\FMMO-\*\]Jc"676RRTUZ[\]h\iZiTjjlmqrxmylz{|#1   7- 113 2 3 HHQKs;',LW\\^,L)LM]+S0n]d]k]k]m0n-no 8} [)3.eT[TbTbTd.e+ef +l:ZG"L  cK  QxjL)M!J<+H8TUVbUcScJddfgrfstu cK !( Iv::mR0"5DVZZ"56J::k2.D Yq l+2d1ZL.-AJqQ]P^N^C__abfagh i "1   7"112d6G6G6M2NMY_c_r_rs 'E KK,_,==PQVPWWefjfyfyez{ |  - -e 4(  fTYY'9:;;TUV  wt]B78K    KK$_$55TUYU^U^_jUkTlmn32#22**2?? 2 (( 2 ** 2!% 2 22!% 2 22 ++2++2$$2((2&&2,,2002 &&!2"**#2$ 22%2&..'2(,,)2,-- 11''--"55  +2:,,..,,,,#'#>#>&*&D&D $ 8 8 92L!44((** $ < <  K2X!22..((+/+P+P00  W2Il  3 37b bkok|k|../43D3D3JK  1 $-a0OO< ))+4KcKc)d  PQ   KK78N7OPQ  *1#. >  sz z= )z88z=)r rrros importlibrr&rrr r)s ~Rr