o rri%@sddlZddlZddlmZddlZddlZddlZddl Z ddl Z ddl Z ddlmZddZGdddZd*d d Zd d ZddZddZddZd+ddZd,ddZ     d-de jde jeefded ed!e jd"d#e jd$d%ed&e jd'fd(d)ZdS).N)Path)temp_filenamesc Cs.tdddt|ddddg}t|dS) Nffprobe -loglevelpanicz -print_formatjsonz -show_formatz -show_streamszutf-8)sp check_outputstrrloadsdecode)path stdout_datar@/home/ubuntu/.local/lib/python3.10/site-packages/demucs/audio.py _read_infos rc@s~eZdZdZdefddZddZeddZed d Z ed d Z d dZ dddZ dddZ ddedddfddZdS) AudioFilez Allows to read audio from any format supported by ffmpeg, as well as resampling or converting to mono on the fly. See :method:`read` for more details. rcCst||_d|_dSN)rr_info)selfrrrr__init__!s  zAudioFile.__init__cCsbd|jfg}|d|f|d|f|dt|fddd|D}d|d S) Nr sampleratechannelsstreamsz, css"|] \}}|d|VqdS)=Nr).0namevaluerrr *s z%AudioFile.__repr__..z AudioFile())rappendrrlenjoin)rfeatures features_strrrr__repr__%s  zAudioFile.__repr__cCs|jdur t|j|_|jSr)rrrrrrrinfo-s  zAudioFile.infocCst|jddS)Nformatduration)floatr(r'rrrr*3szAudioFile.durationcCsddt|jdDS)NcSs g|] \}}|ddkr|qS) codec_typeaudior)rindexstreamrrr 9s  z,AudioFile._audio_streams..r) enumerater(r'rrr_audio_streams7s zAudioFile._audio_streamscCs t|jSr)r"r2r'rrr__len__>s zAudioFile.__len__rcCt|jd|j|dS)Nrrintr(r2rr/rrrrAzAudioFile.channelscCr4)Nr sample_rater5r7rrrrDr8zAudioFile.samplerateNcCsttt||}t|tj }|r|g}|dur d}d}nt|p&||}t|d|p3|}t t|} ddg} | ddg7} |rQ| dt |g7} | dt |j g7} t || D]9\} } | d d |j | g7} |dur{| d t |g7} | d d g7} | ddg7} |dur| dt |g7} | | g7} q_tj| ddg} | D]3} tj| tjd}t|}|d|}|durt||}|dur|dd|f}| |qWdn1swYtj| dd}|r|d}|S)a, Slightly more efficient implementation than stempeg, in particular, this will extract all stems at once rather than having to loop over one file multiple times for each stream. Args: seek_time (float): seek time in seconds or None if no seeking is needed. duration (float): duration in seconds to extract or None to extract until the end. streams (slice, int or list): streams to extract, can be a single int, a list or a slice. If it is a slice or list, the output will be of size [S, C, T] with S the number of streams, C the number of channels and T the number of samples. If it is an int, the output will be [C, T]. samplerate (int): if provided, will resample on the fly. If None, no resampling will be done. Original sampling rate can be obtained with :method:`samplerate`. channels (int): if 1, will convert to mono. We do not rely on ffmpeg for that as ffmpeg automatically scale by +3dB to conserve volume when playing on speakers. See https://sound.stackexchange.com/a/42710. Our definition of mono is simply the average of the two channels. Any other value will be ignored. Nrffmpegz-yrrz-ssz-iz-mapz0:z-tz-threads1z-ff32lez-arT)check)dtype.r)dim)nparrayranger" isinstancendarrayr6rr+rr rzipr2r runfromfilefloat32torch from_numpyviewrtconvert_audio_channelsr!stack)r seek_timer*rrrsingle target_sizequery_duration filenamescommandr/filenamewavswavrrrreadGsR       zAudioFile.read)r)__name__ __module__ __qualname____doc__rrr&propertyr(r*r2r3rrslicerYrrrrrs&     rcCs|j^}}}||kr |S|dkr|jddd}|S|dkr-|jg|||R}|S||kr>|dd|ddf}|Std)z.Convert audio to the given number of channels.rT)r@keepdim.Nz@The audio file has less channels than requested but is not mono.)shapemeanexpand ValueError)rXrrc src_channelslengthrrrrNs rNcCst||}t|||S)zTConvert audio from a given samplerate to a target one and target number of channels.)rNjulius resample_frac)rXfrom_samplerate to_sampleraterrrr convert_audios rmcCs |jjr|dddS|S)z,Convert audio to 16 bits integer PCM format.r?r)r>is_floating_pointclamp_shortrXrrri16_pcmsrscCs|jjr|S|dS)z*Convert audio to float 32 bits PCM format.rn)r>ror+rrrrrf32_pcms rtcCs|jjrt|St|S)zHConvert audio to either f32 pcm or i16 pcm depending on the given dtype.)r>rortrs)rXr>rrr as_dtype_pcmsruD@Fc Cs|j\}}t|}t}|||||||||s'||j }| dd }| |} | |7} t|d} | | WddS1sWwYdS)z6Save given audio as mp3. This should work on all OSes.rrwbN)rcrslameencEncoder set_bit_rateset_in_sample_rate set_channels set_qualitysilencedatacpu transposenumpyencodetobytesflushopenwrite) rXrrbitratequalityverboseCTencodermp3_datafrrr encode_mp3s         "rrescalecCs|dus|dkr |S|jjsJd|dkr%|td|d}|S|dkr1|dd }|S|d krromaxabsrrJrrf)rXmoderrr prevent_clips  rrXrrrclip)rrrrbits_per_sample)r as_floatpreset)r`c Cst||d}t|}|j}|dkrt|||||ddd S|dkr:|r*d}d} nd} tjt|||| |d d S|d krKtjt||||d d Std |)a Save audio file, automatically preventing clipping if necessary based on the given `clip` strategy. If the path ends in `.mp3`, this will save as mp3 with the given `bitrate`. Use `preset` to set mp3 quality: 2 for highest quality, 7 for fastest speed )rz.mp3T)rz.wavrPCM_FPCM_S)r9encodingrz.flac)r9rzInvalid suffix for path: N) rrsuffixlowerrtasaver rf) rXrrrrrrrrrrrr save_audios   r)r`)rvrwr`F)r)rwrrFr`) r subprocessr pathlibrryrirrArJ torchaudiortypingtputilsrrrrNrmrsrtrurrTensorUnionr r6LiteralboolrrrrrsN   s