o
    ±ƒðj©M  ã                   @   sF  U d Z ddlZddlZddlZddlZddlZddlZddlZddlm	Z	 ddl
mZ ddlmZmZmZmZmZmZ ddlZddlZddlmZmZmZ ddlmZmZmZ dd	lmZ dd
l m!Z!m"Z" ddl#m$Z$ da%ee" e&d< e '¡ Z(eƒ Z)dZ*e +d¡Z,e -e.¡Z/e	G dd„ dƒƒZ0e	G dd„ dƒƒZ1e	G dd„ dƒƒZ2dS )z&Phonemization and synthesis for Piper.é    N)Ú	dataclass)ÚPath)ÚAnyÚIterableÚOptionalÚSequenceÚTupleÚUnioné   )ÚPhonemeTypeÚPiperConfigÚSynthesisConfig)ÚBOSÚEOSÚPAD©Úphonemes_to_ids)ÚESPEAK_DATA_DIRÚEspeakPhonemizer)ÚTashkeelDiacritizerÚ_ESPEAK_PHONEMIZERg    Àÿß@z(\[\[.*?\]\])c                   @   s*   e Zd ZU eed< ee ed< eed< dS )ÚPhonemeAlignmentÚphonemeÚphoneme_idsÚnum_samplesN)Ú__name__Ú
__module__Ú__qualname__ÚstrÚ__annotations__r   Úint© r!   r!   ú�/tmp/claude-1000/-var-www-html-academy/3f42efd1-c2b7-4d08-9b05-30d86bb35479/scratchpad/piper/venv/lib/python3.10/site-packages/piper/voice.pyr   !   s   
 r   c                   @   sÞ   e Zd ZU dZeed< 	 eed< 	 eed< 	 ejed< 	 ee	 ed< 	 ee ed< 	 dZ
eej ed	< 	 dZeee  ed
< 	 dZeej ed< dZee ed< dZeee  ed< edejfdd„ƒZedefdd„ƒZdS )Ú
AudioChunkzChunk of raw audio.Úsample_rateÚsample_widthÚsample_channelsÚaudio_float_arrayÚphonemesr   NÚphoneme_id_samplesÚphoneme_alignmentsÚ_audio_int16_arrayÚ_audio_int16_bytesÚ_phoneme_alignmentsÚreturnc                 C   s0   | j du rt | jt t t¡ tj¡| _ | j S )zg
        Get audio as an int16 numpy array.

        :return: Audio data as int16 numpy array.
        N)r+   ÚnpÚclipr'   Ú_MAX_WAV_VALUEÚastypeÚint16©Úselfr!   r!   r"   Úaudio_int16_arrayM   s   
ÿþzAudioChunk.audio_int16_arrayc                 C   s
   | j  ¡ S )zl
        Get audio as 16-bit PCM bytes.

        :return: Audio data as signed 16-bit sample bytes.
        )r6   Útobytesr4   r!   r!   r"   Úaudio_int16_bytes[   s   
zAudioChunk.audio_int16_bytes)r   r   r   Ú__doc__r    r   r/   ÚndarrayÚlistr   r)   r   r*   r   r+   r,   Úbytesr-   Úpropertyr6   r8   r!   r!   r!   r"   r#   (   s2   
 
r#   c                   @   s¨  e Zd ZU dZejed< 	 eed< 	 eZ	e
ed< 	 e
 ¡ Ze
ed< 	 dZeed< dZee ed	< d
Zee ed< eddeddfdeee
f deeee
f  dedeee
f deeee
f  dedd fdd„ƒZdedeee  fdd„Zdee dee fdd„Z		d$dedee dedee fdd„Z			d%dede j!dee dededeee"  fdd „Z#		d$d!ee dee dedee$j%e&e$j%ee$j% f f fd"d#„Z'dS )&Ú
PiperVoicezA voice for Piper.ÚsessionÚconfigÚespeak_data_dirÚdownload_dirTÚuse_tashkeelNÚtashkeel_diacritiziergš™™™™™é?Útaskeen_thresholdFÚ
model_pathÚconfig_pathÚuse_cudaÚinclude_alignmentsr.   c              
   C   sR  |du r| › d�}t  d|¡ t|ddd��}t |¡}W d  ƒ n1 s&w   Y  |r:ddd	ifg}t  d
¡ ndg}|du rEt ¡ }t| ƒ}	|r’z!ddl}
ddl	m
} |
 t| ƒ¡}||ƒ}| ¡ }	t  d|¡ W n% tyz   t  d¡ Y n ty‘ } zt  d|¡ W Y d}~nd}~ww tt |¡tj|	t ¡ |d�t|ƒt|ƒd�S )aÒ  
        Load an ONNX model and config.

        :param model_path: Path to ONNX voice model.
        :param config_path: Path to JSON voice config (defaults to model_path + ".json").
        :param use_cuda: True if CUDA (GPU) should be used instead of CPU.
        :param espeak_data_dir: Path to espeak-ng data dir (defaults to internal data).
        :param download_dir: Path to download resources (defaults to current directory).
        :param include_alignments: If True, patch the model in memory (requires the
            onnx package) so phoneme/audio alignments are available even if the model
            file has not been patched with piper.patch_voice_with_alignment.
        :return: Voice object.
        Nz.jsonzGuessing voice config path: %sÚrzutf-8)ÚencodingÚCUDAExecutionProviderÚcudnn_conv_algo_searchÚ	HEURISTICz
Using CUDAÚCPUExecutionProviderr   r
   )Úadd_alignment_outputz2Patched model in memory for alignments (tensor=%s)zfThe onnx package is required for include_alignments. Install it with: pip install piper-tts[alignment]z%Not patching model for alignments: %s)Úsess_optionsÚ	providers)r@   r?   rA   rB   )Ú_LOGGERÚdebugÚopenÚjsonÚloadr   Úcwdr   ÚonnxÚpatch_voice_with_alignmentrP   ÚSerializeToStringÚImportErrorÚwarningÚ
ValueErrorr>   r   Ú	from_dictÚonnxruntimeÚInferenceSessionÚSessionOptions)rF   rG   rH   rA   rB   rI   Úconfig_fileÚconfig_dictrR   Úmodel_or_pathrY   rP   ÚmodelÚtensor_nameÚerrorr!   r!   r"   rW   z   sX   
ÿþÿÿÿ€þýøzPiperVoice.loadÚtextc              	   C   s
  | j jtjkrtt d|¡ƒgS | j jtjkr9ddlm	} t
| ddƒ}|du r4|| jd ƒ}t| d|ƒ | |¡S | j jtjkr^ddlm} t
| ddƒ}|du rY|ƒ }t| d|ƒ | |¡S | j jtjkrƒdd	lm} t
| d
dƒ}|du r~|ƒ }t| d
|ƒ | |¡S | j jtjkr¨ddlm} t
| ddƒ}|du r£|ƒ }t| d|ƒ | |¡S | j jtjkr¸td| j j› �ƒ‚g }t |¡}d}	t|ƒD ]±\}
}| d¡�rd}	|sØ| g ¡ |
dkrì||
d   d¡rì|d  d¡ |d  |dd…   ¡ ¡ |
t!|ƒd k �r||
d   d¡�r|d  d¡ qÅ| j j"dk�r1| j#�r1| j$du �r)t%ƒ | _$| j$|| j&d�}t'�8 t(du �r>t)| j*ƒa(t(j| j j"|| j j+d�}|	�r_|�r_|d  |d ¡ |dd… }| |¡ W d  ƒ n	1 �sow   Y  d}	qÅ|�rƒ|d �sƒ| ,¡  |S )z•
        Text to phonemes grouped by sentence.

        :param text: Text to phonemize.
        :return: List of phonemes for each sentence.
        ÚNFDr
   )ÚChinesePhonemizerÚ_chinese_phonemizerNÚg2pW)ÚHebrewPhonemizerÚ_hebrew_phonemizer)ÚJapanesePhonemizerÚ_japanese_phonemizer)ÚThaiPhonemizerÚ_thai_phonemizerzUnexpected phoneme type: Fz[[Tr   ú éÿÿÿÿé   éþÿÿÿÚar)rE   )Úvowel_clusters)-r@   Úphoneme_typer   ÚTEXTr;   ÚunicodedataÚ	normalizeÚPINYINÚphonemize_chineserk   ÚgetattrrB   ÚsetattrÚ	phonemizeÚHEBREWÚphonemize_hebrewrn   ÚJAPANESEÚphonemize_japaneserp   ÚTHAIÚphonemize_thairr   ÚESPEAKr^   Ú_PHONEME_BLOCK_PATTERNÚsplitÚ	enumerateÚ
startswithÚappendÚendswithÚextendÚstripÚlenÚespeak_voicerC   rD   r   rE   Ú_ESPEAK_PHONEMIZER_LOCKr   r   rA   ry   Úpop)r5   ri   rk   Ú
phonemizerrn   rp   rr   r(   Ú
text_partsÚprev_raw_phonemesÚiÚ	text_partÚtext_part_phonemesr!   r!   r"   r‚   Ê   sŠ   	





&ÿ

ýñzPiperVoice.phonemizer(   c                 C   s6   | j jtjkrddlm} ||| j jƒS t|| j jƒS )zt
        Phonemes to ids.

        :param phonemes: List of phonemes.
        :return: List of phoneme ids.
        r
   r   )r@   rz   r   r~   r   r   Úphoneme_id_map)r5   r(   Úchinese_phonemes_to_idsr!   r!   r"   r   >  s   zPiperVoice.phonemes_to_idsÚ
syn_configc                 c   sà  � |du rt }|  |¡}t d||¡ |D ]Ø}|sq|  |¡}d}| j|||d�}t|tƒr3|\}	}n|}	|jrNt	 
t	 |	¡¡}
|
dk rJt	 |	¡}	n|	|
 }	|jdkrX|	|j }	t	 |	dd¡ t	j¡}	d}|durÞt|ƒt|ƒkrÞ| jj tg ¡}d}g }d}t tg|tg¡D ]L}| jj |g ¡}|tkrŸtt ||¡ƒ}n|}|}|D ]}|t|ƒkr±d	} n||| kr»d	} n|d
7 }q¥|rÄ n| t||t|||… ƒd�¡ qˆ|rÞd}t d¡ t| jjdd
|	||||d�V  qdS )a  
        Synthesize one audio chunk per sentence from from text.

        :param text: Text to synthesize.
        :param syn_config: Synthesis configuration.
        :param include_alignments: If True and the model supports it, include phoneme/audio alignments.
        Nztext=%s, phonemes=%s)rI   g:Œ0âŽyE>g      ð?g      ð¿r   FTr
   )r   r   r   zPhoneme alignment failedrv   )r$   r%   r&   r'   r(   r   r)   r*   ) Ú_DEFAULT_SYNTHESIS_CONFIGr‚   rS   rT   r   Úphoneme_ids_to_audioÚ
isinstanceÚtupleÚnormalize_audior/   ÚmaxÚabsÚ
zeros_likeÚvolumer0   r2   Úfloat32r’   r@   rœ   Úgetr   Ú	itertoolsÚchainr   r   r;   rŽ   r   Úsumr#   r$   )r5   ri   rž   rI   Úsentence_phonemesr(   r   r)   Úaudio_resultÚaudioÚmax_valr*   Úpad_idsÚphoneme_id_idxÚalignment_failedr   Úexpected_idsÚids_to_checkÚstart_phoneme_id_idxÚ
phoneme_idr!   r!   r"   Ú
synthesizeM  sŒ   €

ÿ





ÿýÿ


ø¯zPiperVoice.synthesizeÚwav_fileÚset_wav_formatc           	      C   s|   g }d}| j |||d�D ]+}|r&|r$| |j¡ | |j¡ | |j¡ d}| |j¡ |r7|j	r7| 
|j	¡ q|r<|S dS )a×  
        Synthesize and write WAV audio from text.

        :param text: Text to synthesize.
        :param wav_file: WAV file writer.
        :param syn_config: Synthesis configuration.
        :param set_wav_format: True if the WAV format should be set automatically.
        :param include_alignments: If True and the model supports it, return phoneme/audio alignments.

        :return: Phoneme/audio alignments if include_alignments is True, otherwise None.
        T)rž   rI   FN)r¸   Úsetframerater$   Úsetsampwidthr%   Úsetnchannelsr&   Úwriteframesr8   r*   r�   )	r5   ri   r¹   rž   rº   rI   Ú
alignmentsÚfirst_chunkÚaudio_chunkr!   r!   r"   Úsynthesize_wav¼  s$   
ÿ
€zPiperVoice.synthesize_wavr   c                 C   sN  |du rt }|j}|j}|j}|j}|du r| jj}|du r"| jj}|du r*| jj}t tj|tj	d�d¡}tj|j
d gtj	d�}	tj|||gtjd�}
||	|
dœ}| jjdkr[d}| jjdkri|du ri| jj}|durztj|gtj	d�}||d< | j d|¡}|d  ¡ }|s‹|S t|ƒdkr•|dfS |d  ¡ | jj  tj	¡}||fS )a4  
        Synthesize raw audio from phoneme ids.

        :param phoneme_ids: List of phoneme ids.
        :param syn_config: Synthesis configuration.
        :param include_alignments: Return samples per phoneme id if True.
        :return: Audio float numpy array from voice model (unnormalized, in range [-1, 1]).

        If include_alignments is True and the voice model supports it, the return
        value will be a tuple instead with (audio, phoneme_id_samples) where
        phoneme_id_samples contains the number of audio samples per phoneme id.
        N)Údtyper   r
   )ÚinputÚinput_lengthsÚscalesÚsid)rŸ   Ú
speaker_idÚlength_scaleÚnoise_scaleÚnoise_w_scaler@   r/   Úexpand_dimsÚarrayÚint64Úshaper¨   Únum_speakersÚdefault_speaker_idr?   ÚrunÚsqueezer’   Ú
hop_lengthr2   )r5   r   rž   rI   rÈ   rÉ   rÊ   rË   Úphoneme_ids_arrayÚphoneme_ids_lengthsrÆ   ÚargsrÇ   Úresultr¯   r)   r!   r!   r"   r    ç  sT   þýþÿzPiperVoice.phoneme_ids_to_audio)NF)NTF)(r   r   r   r9   r`   ra   r   r   r   rA   r   rX   rB   rC   ÚboolrD   r   r   rE   ÚfloatÚstaticmethodr	   r   rW   r;   r‚   r    r   r   r   r#   r¸   ÚwaveÚ
Wave_writer   rÂ   r/   r:   r   r    r!   r!   r!   r"   r>   e   s”   
 
ú
ÿþý
üûúùOtüþýü
ûsúþýüûú

ù.üþýüûr>   )3r9   rª   rV   ÚloggingÚreÚ	threadingr|   rÜ   Údataclassesr   Úpathlibr   Útypingr   r   r   r   r   r	   Únumpyr/   r`   r@   r   r   r   Úconstr   r   r   r   r   Úphonemize_espeakr   r   Útashkeelr   r   r   ÚLockr”   rŸ   r1   ÚcompilerŠ   Ú	getLoggerr   rS   r   r#   r>   r!   r!   r!   r"   Ú<module>   s<     

<