o jS\@sddlmZddlmZmZddlZddlZddlZ ddl Z ddl m Z ddl mZmZmZmZmZmZmZmZmZmZmZmZmZmZmZmZGdddeZdS))BytesIO)DictTupleN)StandardScaler) amp_to_dbbuild_mel_basis compute_f0 db_to_amp deemphasis find_endpoint griffin_limload_wav mel_to_specmillisec_to_length preemphasisrms_volume_norm spec_to_melstft trim_silence volume_normc @s<eZdZdZ                             dKd d ZedLdMddZdejdejfddZ dejdejfddZ de de ej ej ej ej effddZdejdejdejdejddf ddZd ejdejfd!d"Zd ejdejfd#d$Zd%ejdejfd&d'Zd%ejdejfd(d)Zd*ejdejfd+d,Zd-ejdejfd.d/Zd0ejdejfd1d2Zd3d4Zd ejdejfd5d6ZdNd8ejdefd9d:Zd;d<Zed ejdejfd=d>ZdOd ejd?edejfd@dAZdOdBe dCedejfdDdEZ dPd8ejdFe dCeddfdGdHZ!dBe defdIdJZ"dS)QAudioProcessoraAudio Processor for TTS. Note: All the class arguments are set to default values to enable a flexible initialization of the class with the model config. They are not meaningful for all the arguments. Args: sample_rate (int, optional): target audio sampling rate. Defaults to None. resample (bool, optional): enable/disable resampling of the audio clips when the target sampling rate does not match the original sampling rate. Defaults to False. num_mels (int, optional): number of melspectrogram dimensions. Defaults to None. log_func (int, optional): log exponent used for converting spectrogram aplitude to DB. min_level_db (int, optional): minimum db threshold for the computed melspectrograms. Defaults to None. frame_shift_ms (int, optional): milliseconds of frames between STFT columns. Defaults to None. frame_length_ms (int, optional): milliseconds of STFT window length. Defaults to None. hop_length (int, optional): number of frames between STFT columns. Used if ```frame_shift_ms``` is None. Defaults to None. win_length (int, optional): STFT window length. Used if ```frame_length_ms``` is None. Defaults to None. ref_level_db (int, optional): reference DB level to avoid background noise. In general <20DB corresponds to the air noise. Defaults to None. fft_size (int, optional): FFT window size for STFT. Defaults to 1024. power (int, optional): Exponent value applied to the spectrogram before GriffinLim. Defaults to None. preemphasis (float, optional): Preemphasis coefficient. Preemphasis is disabled if == 0.0. Defaults to 0.0. signal_norm (bool, optional): enable/disable signal normalization. Defaults to None. symmetric_norm (bool, optional): enable/disable symmetric normalization. If set True normalization is performed in the range [-k, k] else [0, k], Defaults to None. max_norm (float, optional): ```k``` defining the normalization range. Defaults to None. mel_fmin (int, optional): minimum filter frequency for computing melspectrograms. Defaults to None. mel_fmax (int, optional): maximum filter frequency for computing melspectrograms. Defaults to None. pitch_fmin (int, optional): minimum filter frequency for computing pitch. Defaults to None. pitch_fmax (int, optional): maximum filter frequency for computing pitch. Defaults to None. spec_gain (int, optional): gain applied when converting amplitude to DB. Defaults to 20. stft_pad_mode (str, optional): Padding mode for STFT. Defaults to 'reflect'. clip_norm (bool, optional): enable/disable clipping the our of range values in the normalized audio signal. Defaults to True. griffin_lim_iters (int, optional): Number of GriffinLim iterations. Defaults to None. do_trim_silence (bool, optional): enable/disable silence trimming when loading the audio signal. Defaults to False. trim_db (int, optional): DB threshold used for silence trimming. Defaults to 60. do_sound_norm (bool, optional): enable/disable signal normalization. Defaults to False. do_amp_to_db_linear (bool, optional): enable/disable amplitude to dB conversion of linear spectrograms. Defaults to True. do_amp_to_db_mel (bool, optional): enable/disable amplitude to dB conversion of mel spectrograms. Defaults to True. do_rms_norm (bool, optional): enable/disable RMS volume normalization when loading an audio file. Defaults to False. db_level (int, optional): dB level used for rms normalization. The range is -99 to 0. Defaults to None. stats_path (str, optional): Path to the computed stats file. Defaults to None. verbose (bool, optional): enable/disable logging. Defaults to True. NFnp.log10reflectT<c"*Ks||_||_||_||_|pd|_||_||_| |_| |_| |_ | |_ ||_ ||_ ||_ |p/d|_||_||_||_t||_||_|durHdnt||_||_||_||_||_||_||_||_||_| |_|dkrqtj |_!n |dkryd|_!nt"d|durt#|j|j|jd\|_$|_%n||_%| |_$|dksJd |j$|jksJd |j$d |jt&|}#|!rt'd |#(D] \}$}%t'd )|$|%qt*|j|j|j|j|jd|_+| r|r|,| \}&}'}(})}"|-|&|'|(|)d|_ d|_d|_d|_ dSdSdS)Nrg?znp.logr z [!] unknown `log_func` value.)frame_length_msframe_shift_ms sample_raterz [!] min_level_db is 0z1 [!] win_length cannot be larger than fft_size -  vs z > Setting up Audio Processor...z | > {}:{})r fft_sizenum_melsmel_fmaxmel_fminT).r resampler#log_func min_level_dbrr ref_level_dbr"powerrgriffin_lim_iters signal_normsymmetric_normr%r$ pitch_fmin pitch_fmaxfloat spec_gain stft_pad_modemax_norm clip_normdo_trim_silencetrim_db do_sound_normdo_amp_to_db_lineardo_amp_to_db_mel do_rms_normdb_level stats_pathnpebase ValueErrorr win_length hop_lengthvarsprintitemsformatr mel_basis load_stats setup_scaler)*selfr r&r#r'r(rrrBrAr)r"r*rr,r-r3r%r$r/r.r1r2r4r+r5r6r7r8r9r:r;r<verbose_memberskeyvaluemel_meanmel_std linear_mean linear_stdrTL/home/kuhnn/.local/lib/python3.10/site-packages/TTS/utils/audio/processor.py__init__s&       zAudioProcessor.__init__configCoqpitcCs.d|vrtdd|i|jStdd|i|S)NaudiorKrT)rrY)rWrKrTrTrUinit_from_configszAudioProcessor.init_from_configSreturncCs|}|jrnt|dr2|jd|jkr|j|jjS|jd|jdkr.|j |jjSt d||j 8}||j |j }|j r\d|j||j}|jrZt||j |j}|S|j|}|jrlt|d|j}|S|S)aFNormalize values into `[0, self.max_norm]` or `[-self.max_norm, self.max_norm]` Args: S (np.ndarray): Spectrogram to normalize. Raises: RuntimeError: Mean and variance is computed from incompatible parameters. Returns: np.ndarray: Normalized spectrogram. mel_scalerr@ [!] Mean-Var stats does not match the given feature dimensions.)copyr,hasattrshaper#r] transformTr" linear_scaler RuntimeErrorr)r(r-r3r4r=clip)rJr[S_normrTrTrU normalizes,     zAudioProcessor.normalizecCs|}|jrtt|dr2|jd|jkr|j|jjS|jd|jdkr.|j |jjSt d|j rX|j rBt ||j |j}||j|j d|j|j}||jS|j rct |d|j}||j |j|j}||jS|S)aDenormalize spectrogram values. Args: S (np.ndarray): Spectrogram to denormalize. Raises: RuntimeError: Mean and variance are incompatible. Returns: np.ndarray: Denormalized spectrogram. r]rr^r_)r`r,rarbr#r]inverse_transformrdr"rerfr-r4r=rgr3r(r))rJr[S_denormrTrTrU denormalize,s(   "  zAudioProcessor.denormalizer<c Cstj|dd}|d}|d}|d}|d}|d}gd}|D]&} | |vr,q%| d vrK|| |j| ksKJd | d || d |j| q%|||||fS) a)Loading mean and variance statistics from a `npy` file. Args: stats_path (str): Path to the `npy` file containing Returns: Tuple[np.array, np.array, np.array, np.array, Dict]: loaded statistics and the config used to compute them. T) allow_picklerPrQrRrS audio_config)r+r<r5r)r*)r r6z [!] Audio param z= does not match the value used for computing mean-var stats. r!)r=loaditemkeys__dict__) rJr<statsrPrQrRrS stats_configskip_parametersrNrTrTrUrHSs   zAudioProcessor.load_statsrPrQrRrScCs0t|_|j||t|_|j||dS)aWInitialize scaler objects used in mean-std normalization. Args: mel_mean (np.ndarray): Mean for melspectrograms. mel_std (np.ndarray): STD for melspectrograms. linear_mean (np.ndarray): Mean for full scale spectrograms. linear_std (np.ndarray): STD for full scale spectrograms. N)rr] set_statsre)rJrPrQrRrSrTrTrUrIos zAudioProcessor.setup_scalerxcCst||jdS)a=Apply pre-emphasis to the audio signal. Useful to reduce the correlation between neighbouring signal values. Args: x (np.ndarray): Audio signal. Raises: RuntimeError: Preemphasis coeff is set to 0. Returns: np.ndarray: Decorrelated audio signal. rwcoef)rrJrwrTrTrUapply_preemphasiss z AudioProcessor.apply_preemphasiscCst||jdS)zReverse pre-emphasis.rx)r rrzrTrTrUapply_inv_preemphasissz$AudioProcessor.apply_inv_preemphasisycCsj|jdkr ||}t||j|j|j|jd}|jr'tt ||j |j d}nt |}| |t jS)zCompute a spectrogram from a waveform. Args: y (np.ndarray): Waveform. Returns: np.ndarray: Spectrogram. rr}r"rBrApad_moderwgainr?)rr{rr"rBrAr2r8rr=absr1r?riastypefloat32rJr}Dr[rTrTrU spectrograms  zAudioProcessor.spectrogramcCsl|jdkr ||}t||j|j|j|jd}tt ||j d}|j r-t ||j |jd}||tjS)z)Compute a melspectrogram from a waveform.rr~specrGr)rr{rr"rBrAr2rr=rrGr9rr1r?rirrrrTrTrUmelspectrograms  zAudioProcessor.melspectrogramrcCsD||}t||j|jd}|||j}|jdkr ||S|S)z=Convert a spectrogram to a waveform using Griffi-Lim vocoder.rr)rlr r1r? _griffin_limr*rr|)rJrr[WrTrTrUinv_spectrograms zAudioProcessor.inv_spectrogrammel_spectrogramcCsR||}t||j|jd}t||jd}|||j}|jdkr'| |S|S)z@Convert a melspectrogram to a waveform using Griffi-Lim vocoder.r)melrGr) rlr r1r?rrGrr*rr|)rJrrr[rrTrTrUinv_melspectrograms z!AudioProcessor.inv_melspectrogram linear_speccCsP||}t||j|jd}tt||jd}t||j|jd}| |}|S)zConvert a full scale linear spectrogram output of a network to a melspectrogram. Args: linear_spec (np.ndarray): Normalized full scale linear spectrogram. Returns: np.ndarray: Normalized melspectrogram. rr) rlr r1r?rr=rrGrri)rJrr[rrTrTrUout_linear_to_mels  z AudioProcessor.out_linear_to_melcCst||j|j|j|j|jdS)N)rnum_iterrBrAr"r)r r+rBrAr"r2)rJr[rTrTrUrszAudioProcessor._griffin_limc CsVt||jdkrtj|d|jdf|jd}t||j|j|j|j|j |jdd}|S)aCompute pitch (f0) of a waveform using the same parameters used for computing melspectrogram. Args: x (np.ndarray): Waveform. Returns: np.ndarray: Pitch. Examples: >>> WAV_FILE = filename = librosa.example('vibeace') >>> from TTS.config import BaseAudioConfig >>> from TTS.utils.audio import AudioProcessor >>> conf = BaseAudioConfig(pitch_fmax=640, pitch_fmin=1) >>> ap = AudioProcessor(**conf) >>> wav = ap.load_wav(WAV_FILE, sr=ap.sample_rate)[:5 * ap.sample_rate] >>> pitch = ap.compute_f0(wav) rr^)modeT)rwr/r.rBrAr r2center) lenrBr=padr2rr/r.rAr )rJrwf0rTrTrUrs zAudioProcessor.compute_f0皙?wavcCst||j|j||j|jdS)aFind the last point without silence at the end of a audio signal. Args: wav (np.ndarray): Audio signal. threshold_db (int, optional): Silence threshold in decibels. Defaults to -40. min_silence_sec (float, optional): Ignore silences that are shorter then this in secs. Defaults to 0.8. Returns: int: Last point without silence. )rr6r min_silence_secrr?)r r6r r1r?)rJrrrTrTrUr s zAudioProcessor.find_endpointcCst||j|j|j|jdS)z6Trim silent parts with a threshold and 0.01 sec margin)rr r6rArB)rr r6rArB)rJrrTrTrUrszAudioProcessor.trim_silencecCs t|dS)zNormalize the volume of an audio signal. Args: x (np.ndarray): Raw waveform. Returns: np.ndarray: Volume normalized waveform. rw)rrrTrTrU sound_norm(s zAudioProcessor.sound_normr;cCs|dur|j}t||dS)zNormalize the volume based on RMS of the signal. Args: x (np.ndarray): Raw waveform. Returns: np.ndarray: RMS normalized waveform. N)rwr;)r;r)rJrwr;rTrTrUr4s  zAudioProcessor.rms_volume_normfilenamesrcCs|dur t||dd}n t||j|jd}|jr0z||}Wnty/td|Ynw|jr8||}|j rB| ||j }|S)aRead a wav file using Librosa and optionally resample, silence trim, volume normalize. Resampling slows down loading the file significantly. Therefore it is recommended to resample the file before. Args: filename (str): Path to the wav file. sr (int, optional): Sampling rate for resampling. Defaults to None. Returns: np.ndarray: Loaded waveform. NT)rr r&z* [!] File cannot be trimmed for silence - ) r r r&r5rr@rDr7rr:rr;)rJrrrwrTrTrUr Bs   zAudioProcessor.load_wavpathc Cs|jr |||jd}n|dtdtt|}|tj}|rBt}t j j ||r0|n|j ||d|j |t j j ||rK|n|j |dS)a^Save a waveform to a file using Scipy. Args: wav (np.ndarray): Waveform to save. path (str): Path to a output file. sr (int, optional): Sampling rate used for saving to the file. Defaults to None. pipe_out (BytesIO, optional): Flag to stdout the generated TTS wav file for shell pipe. ig{Gz?rN)r:rr;maxr=rrint16rscipyiowavfilewriter seekbufferread)rJrrrpipe_outwav_norm wav_bufferrTrTrUsave_wav]s    zAudioProcessor.save_wavcCs tj|dS)zwGet the duration of a wav file using Librosa. Args: filename (str): Path to the wav file. )r)librosa get_duration)rJrrTrTrUrss zAudioProcessor.get_duration)!NFNrNNNNNNrNrNNNNNNNrrTNFrFTTFNNT)T)rWrX)r)N)NN)#__name__ __module__ __qualname____doc__rV staticmethodrZr=ndarrayrirlstrrarrayrrHrIr{r|rrrrrrrintr rrr0rr rrrTrTrTrUr sn o)('  $  r) rrtypingrrrnumpyr=scipy.io.wavfiler scipy.signalTTS.tts.utils.helpersr TTS.utils.audio.numpy_transformsrrrr r r r r rrrrrrrrobjectrrTrTrTrUs  H