+
    ƒ°ñj##  ã                   óø   € ^ RI t ^ RIHt ^ RIt^ RIt^ RIHt ^ RIH	t	 ^ RI
HtHt ^ RIHtHt ^t]! RR7       ! R	 R
4      4       t ! R R4      t ! R R4      t ! R R]P(                  P*                  4      tR# )é    N)Ú	dataclass)ÚDiagonalGaussianDistribution)ÚAudioPatchifier)ÚCausalityAxisÚCausalAudioAutoencoder)ÚVocoderÚVocoderWithBWET)Úfrozenc                   óL   a € ] tR t^t o Rt]V 3R lR l4       tV 3R ltRtV t	R# )ÚAudioVAEComponentConfigzDContainer for model component configuration extracted from metadata.c                ó$   <€ V ^8„  d   QhRS[ RR/# )é   ÚmetadataÚreturnr   ©Údict)ÚformatÚ__classdict__s   "€ÚFC:\Ai_Coding\apps\studio\ComfyUI\comfy\ldm\lightricks\vae\audio_vae.pyÚ__annotate__Ú$AudioVAEComponentConfig.__annotate__   s    ø€ ÷ Eñ E¡Tð EÐ.Gñ Eó    c                ó  € Ve   RV9   g   Q R4       hVR,          p\        V\        4      '       d   \        P                  ! V4      pMTpVP	                  R4      pVP	                  R4      pVf   Q R4       hVf   Q R4       hV ! WER7      # )NÚconfigz"Metadata is required for audio VAEÚ	audio_vaeÚvocoderz*Audio VAE config is required for audio VAEz(Vocoder config is required for audio VAE)Úautoencoderr   )Ú
isinstanceÚstrÚjsonÚloadsÚget)Úclsr   Ú
raw_configÚparsed_configÚaudio_configÚvocoder_configs   &&    r   Úfrom_metadataÚ%AudioVAEComponentConfig.from_metadata   s”   € àÒ#¨°HÔ(<ÐbÐ>bÓbÐ<à˜hÕ'ˆ
Ü�j¤#×&Ò&Ü ŸJšJ zÓ2‰Mà&ˆMà$×(Ñ(¨Ó5ˆØ&×*Ñ*¨9Ó5ˆàÒ'ÐUÐ)UÓUÐ'ØÒ)ÐUÐ+UÓUÐ)á˜|ÔDÐDr   c                ó2   <€ V ^8„  d   Qh/ S[ ;R&   S[ ;R&   # )r   r   r   r   )r   r   s   "€r   r   r      s   ø‡ ‚ ñ Ññ	 ñ
 �Mò r   © N)
Ú__name__Ú
__module__Ú__qualname__Ú__firstlineno__Ú__doc__Úclassmethodr(   Ú__annotate_func__Ú__static_attributes__Ú__classdictcell__©r   s   @r   r   r      s%   ø‡ € áNð
 ÷Eó ðE÷ ƒ r   r   c                   óZ   a € ] tR t^*t o RtV 3R lR ltV 3R lR ltV 3R lR ltRtV t	R	# )
ÚAudioLatentNormalizerzKApplies per-channel statistics in patch space and restores original layout.c                óN   <€ V ^8„  d   QhRS[ RS[P                  P                  /# )r   Ú	patchfierÚstatistics_processor)r   ÚtorchÚnnÚModule)r   r   s   "€r   r   Ú"AudioLatentNormalizer.__annotate__-   s#   ø€ ÷ /ñ /¡/ð /ÉÏÉÏÉñ /r   c                ó   € Wn         W n        R # ©N©Ú
patchifierÚ
statistics)Úselfr9   r:   s   &&&r   Ú__init__ÚAudioLatentNormalizer.__init__-   s   € Ø#ŒØ.Žr   c                óN   <€ V ^8„  d   QhRS[ P                  RS[ P                  /# ©r   Úlatentsr   ©r;   ÚTensor)r   r   s   "€r   r   r>   1   s'   ø€ ÷ Tñ T¡§¡ð T±%·,±,ñ Tr   c                óø   € VP                   ^,          pVP                   ^,          pV P                  P                  V4      w  rEV P                  P	                  V4      pV P                  P                  WbVR7      # ©é   )ÚchannelsÚfreq)ÚshaperB   ÚpatchifyrC   Ú	normalizeÚ
unpatchify)rD   rI   rO   rP   ÚpatchedÚ_Ú
normalizeds   &&     r   rS   ÚAudioLatentNormalizer.normalize1   sb   € Ø—=‘= Õ#ˆØ�}‰}˜QÕˆØ—_‘_×-Ñ-¨gÓ6‰
ˆØ—_‘_×.Ñ.¨wÓ7ˆ
Ø�‰×)Ñ)¨*ÈdÐ)ÓSÐSr   c                óN   <€ V ^8„  d   QhRS[ P                  RS[ P                  /# rH   rJ   )r   r   s   "€r   r   r>   8   s'   ø€ ÷ Vñ V¡5§<¡<ð V±E·L±Lñ Vr   c                óø   € VP                   ^,          pVP                   ^,          pV P                  P                  V4      w  rEV P                  P	                  V4      pV P                  P                  WbVR7      # rM   )rQ   rB   rR   rC   Úun_normalizerT   )rD   rI   rO   rP   rU   rV   Údenormalizeds   &&     r   ÚdenormalizeÚ!AudioLatentNormalizer.denormalize8   sc   € Ø—=‘= Õ#ˆØ�}‰}˜QÕˆØ—_‘_×-Ñ-¨gÓ6‰
ˆØ—‘×3Ñ3°GÓ<ˆØ�‰×)Ñ)¨,ÐPTÐ)ÓUÐUr   rA   N)
r,   r-   r.   r/   r0   rE   rS   r]   r3   r4   r5   s   @r   r7   r7   *   s*   ø‡ € ÙU÷/ð /÷Tð T÷Vö Vr   r7   c                   óZ   a € ] tR t^@t o RtV 3R lR ltV 3R lR ltV 3R lR ltRtV t	R	# )
ÚAudioPreprocessorzKPrepares raw waveforms for the autoencoder by matching training conditions.c                ó2   <€ V ^8„  d   QhRS[ RS[ RS[ RS[ /# )r   Útarget_sample_rateÚmel_binsÚmel_hop_lengthÚn_fft©Úint)r   r   s   "€r   r   ÚAudioPreprocessor.__annotate__C   s*   ø€ ÷ ñ ©3ð ¹#ð Ésð Ñ[^ñ r   c                ó6   € Wn         W n        W0n        W@n        R # r@   ©rb   rc   rd   re   )rD   rb   rc   rd   re   s   &&&&&r   rE   ÚAudioPreprocessor.__init__C   s   € Ø"4ÔØ ŒØ,ÔØŽ
r   c                óT   <€ V ^8„  d   QhRS[ P                  RS[RS[ P                  /# )r   ÚwaveformÚsource_rater   ©r;   rK   rg   )r   r   s   "€r   r   rh   I   s/   ø€ ÷ Tñ T¡§¡ð T¹Cð TÁEÇLÁLñ Tr   c                óz   € W P                   8X  d   V# \        P                  P                  WV P                   4      # r@   )rb   ÚcomfyÚaudioÚresample)rD   rm   rn   s   &&&r   rs   ÚAudioPreprocessor.resampleI   s0   € Ø×1Ñ1Ô1ØˆOÜ�{‰{×#Ñ# H¸4×;RÑ;RÓSÐSr   c                óT   <€ V ^8„  d   QhRS[ P                  RS[RS[ P                  /# )r   rm   Úwaveform_sample_rater   ro   )r   r   s   "€r   r   rh   N   s-   ø€ ÷ 4ñ 4ÙŸ™ð4Ù<?ð4á	�‰ñ4r   c                óÂ  € V P                  W4      p\        P                  P                  V P                  V P
                  V P
                  V P                  R V P                  R,          V P                  RR7      P                  V4      pV! V4      p\        P                  ! \        P                  ! VRR7      4      pVP                  ^ ^^^4      P                  4       # )g        g       @g      ð?)Úsample_ratere   Ú
win_lengthÚ
hop_lengthÚf_minÚf_maxÚn_melsÚpowergñhãˆµøä>)Úmin)rs   rq   rr   ÚMelSpectrogramrb   re   rd   rc   Útor;   ÚlogÚclampÚpermuteÚ
contiguous)rD   rm   rv   ÚdeviceÚmel_transformÚmels   &&&&  r   Úwaveform_to_melÚ!AudioPreprocessor.waveform_to_melN   s·   € ð —=‘= Ó@ˆäŸ™×2Ñ2Ø×/Ñ/Ø—*‘*Ø—z‘zØ×*Ñ*ØØ×)Ñ)¨CÕ/Ø—=‘=Øð 3ó 	
÷ ‰"ˆV‹*ð 	ñ ˜HÓ%ˆÜ�iŠiœŸš C¨TÔ2Ó3ˆØ�{‰{˜1˜a  AÓ&×1Ñ1Ó3Ð3r   )rc   rd   re   rb   N)
r,   r-   r.   r/   r0   rE   rs   r‰   r3   r4   r5   s   @r   r`   r`   @   s(   ø‡ € ÙU÷ð ÷Tð T÷
4ö 4r   r`   c                   ó^  a a€ ] tR t^ct oRtV3R lV 3R lltRV3R lR lltV3R lR ltR tV3R	 lR
 lt	V3R lR lt
]V3R lR l4       t]V3R lR l4       t]V3R lR l4       t]V3R lR l4       t]V3R lR l4       t]V3R lR l4       t]V3R lR l4       tR tRtVtV ;t# )ÚAudioVAEzEHigh-level Audio VAE wrapper exposing encode and decode entry points.c                ó    <€ V ^8„  d   QhRS[ /# )r   r   r   )r   r   s   "€r   r   ÚAudioVAE.__annotate__f   s   ø€ ÷ 
ñ 
¡ñ 
r   c           
     ó@  <€ \         SV `  4        \        P                  V4      p\	        VP
                  R 7      V n        RVP                  9   d   \        VP                  R 7      V n        M\        VP                  R 7      V n        V P
                  P                  4       p\        \        ^\        VR,          VR,          VR,          R7      V P
                  P                  4      V n        \        VR,          VR,          VR,          VR,          R7      V n        R	# )
)r   ÚbweÚsampling_raterd   Ú	is_causal)Ú
patch_sizeÚaudio_latent_downsample_factorrx   rz   r’   rc   re   rj   N)ÚsuperrE   r   r(   r   r   r   r	   r   Ú
get_configr7   r   ÚLATENT_DOWNSAMPLE_FACTORÚper_channel_statisticsÚ
normalizerr`   Úpreprocessor)rD   r   Úcomponent_configÚautoencoder_configÚ	__class__s   &&  €r   rE   ÚAudioVAE.__init__f   sñ   ø€ Ü‰ÑÔä2×@Ñ@ÀÓJÐä1Ð9I×9UÑ9UÔVˆÔØÐ$×,Ñ,Ô,Ü)Ð1A×1IÑ1IÔJˆD�Lä"Ð*:×*BÑ*BÔCˆDŒLà!×-Ñ-×8Ñ8Ó:ÐÜ/ÜØÜ/GØ.¨Õ?Ø-Ð.>Õ?Ø,¨[Õ9ôð ×Ñ×3Ñ3ó	
ˆŒô .Ø1°/ÕBØ'¨
Õ3Ø-Ð.>Õ?Ø$ WÕ-ô	
ˆÖr   c                ó4   <€ V ^8„  d   QhRS[ P                  /# ©r   r   rJ   )r   r   s   "€r   r   rŽ   „   s   ø€ ÷ +ñ +±%·,±,ñ +r   c                óf  € TpTpVP                   pV P                  P                  P                  pVP                  ^,          V8w  db   VP                  ^,          ^8X  d(   VP
                  ! RV.VP                  R,          O5!  pM"\        RV RVP                  ^,           24      hV P                  P                  W4VP                   R7      pV P                  P                  V4      p\        V4      p	V	P                  4       p
V P                  P                  V
4      pVP                  V4      # )z<Encode a waveform dictionary into normalized latent tensors.:r   NNzInput audio must have z channels, got )r†   éÿÿÿÿ)r†   r   ÚencoderÚin_channelsrQ   ÚexpandÚ
ValueErrorrš   r‰   Úencoder   Úmoder™   rS   r�   )rD   rr   rx   rm   rv   Úinput_deviceÚexpected_channelsÚmel_specrI   Ú	posteriorÚlatent_moderW   s   &&&         r   r§   ÚAudioVAE.encode„   s  € ð ˆØ*ÐØ—‘ˆØ ×,Ñ,×4Ñ4×@Ñ@ÐØ�>‰>˜!ÕÐ 1Ô1Ø�~‰~˜aÕ  AÔ%Ø#Ÿ?š?¨2Ð/@ÐVÀ8Ç>Á>ÐRTÕCUÓV‘ä Ø,Ð->Ð,?¸ÈxÏ~É~Ð^_ÕO`ÐNaÐbóð ð ×$Ñ$×4Ñ4Ø°8·?±?ð 5ó 
ˆð ×"Ñ"×)Ñ)¨(Ó3ˆÜ0°Ó9ˆ	Ø—n‘nÓ&ˆà—_‘_×.Ñ.¨{Ó;ˆ
Ø�}‰}˜\Ó*Ð*r   c                óN   <€ V ^8„  d   QhRS[ P                  RS[ P                  /# rH   rJ   )r   r   s   "€r   r   rŽ   ž   s#   ø€ ÷ 
ñ 
™eŸl™lð 
©u¯|©|ñ 
r   c                óÐ   € VP                   pV P                  P                  V4      pV P                  V4      pV P                  P                  WR7      pV P                  V4      pV# )z8Decode normalized latent tensors into an audio waveform.)Útarget_shape)rQ   r™   r]   Útarget_shape_from_latentsr   ÚdecodeÚrun_vocoder)rD   rI   Úoriginal_shaper±   r«   rm   s   &&    r   r³   ÚAudioVAE.decodež   s]   € à Ÿ™ˆà—/‘/×-Ñ-¨gÓ6ˆà×5Ñ5°nÓEˆØ×#Ñ#×*Ñ*¨7Ð*ÓNˆà×#Ñ# HÓ-ˆØˆr   c                ó  € Vw  r#rCV\         ,          pV P                  P                  \        P                  8w  d   V\         ^,
          ,          pVV P                  P
                  P                  VV P                  P                  3# )rN   )r—   r   Úcausality_axisr   ÚNONEÚdecoderÚout_chrc   )rD   Úlatents_shapeÚbatchrV   ÚtimeÚtarget_lengths   &&    r   r²   Ú"AudioVAE.target_shape_from_latentsª   su   € Ø)Ñˆ�$ØÔ7Õ7ˆØ×Ñ×*Ñ*¬m×.@Ñ.@Ô@ØÔ5¸Õ9Õ9ˆMàØ×Ñ×$Ñ$×+Ñ+ØØ×Ñ×%Ñ%ð	
ð 	
r   c                ó,   <€ V ^8„  d   QhRS[ RS[RS[ /# )r   Úframes_numberÚ
frame_rater   )rg   Úfloat)r   r   s   "€r   r   rŽ   ¶   s(   ø€ ÷ Tñ T¹ð TÉð TÑSVñ Tr   c                óZ   € \        \        V4      V,          V P                  ,          4      # r@   )ÚroundrÄ   Úlatents_per_second)rD   rÂ   rÃ   s   &&&r   Únum_of_latents_from_framesÚ#AudioVAE.num_of_latents_from_frames¶   s"   € Ü”e˜MÓ*¨ZÕ7¸4×;RÑ;RÕRÓSÐSr   c                óN   <€ V ^8„  d   QhRS[ P                  RS[ P                  /# )r   r«   r   rJ   )r   r   s   "€r   r   rŽ   ¹   s#   ø€ ÷ 	+ñ 	+¡E§L¡Lð 	+±U·\±\ñ 	+r   c                óä   € V P                   P                  P                  pVP                  ^^4      pV^8X  d   VP	                  ^4      pMV^8w  d   \        RV 24      hV P                  V4      # )r   zUnsupported audio_channels: )r   rº   r»   Ú	transposeÚsqueezer¦   r   )rD   r«   Úaudio_channelsÚvocoder_inputs   &&  r   r´   ÚAudioVAE.run_vocoder¹   sn   € Ø×)Ñ)×1Ñ1×8Ñ8ˆØ ×*Ñ*¨1¨aÓ0ˆà˜QÔØ)×1Ñ1°!Ó4‰MØ˜qÔ ÜÐ;¸NÐ;KÐLÓMÐMà�|‰|˜MÓ*Ð*r   c                ó    <€ V ^8„  d   QhRS[ /# r    rf   )r   r   s   "€r   r   rŽ   Å   s   ø€ ÷ 3ñ 3™Sñ 3r   c                ó@   € \        V P                  P                  4      # r@   )rg   r   r‘   ©rD   s   &r   rx   ÚAudioVAE.sample_rateÄ   s   € ä�4×#Ñ#×1Ñ1Ó2Ð2r   c                ó    <€ V ^8„  d   QhRS[ /# r    rf   )r   r   s   "€r   r   rŽ   É   s   ø€ ÷ 4ñ 4¡ñ 4r   c                ó@   € \        V P                  P                  4      # r@   )rg   r   rd   rÓ   s   &r   rd   ÚAudioVAE.mel_hop_lengthÈ   s   € ä�4×#Ñ#×2Ñ2Ó3Ð3r   c                ó    <€ V ^8„  d   QhRS[ /# r    rf   )r   r   s   "€r   r   rŽ   Í   s   ø€ ÷ .ñ .™#ñ .r   c                ó@   € \        V P                  P                  4      # r@   )rg   r   rc   rÓ   s   &r   rc   ÚAudioVAE.mel_binsÌ   s   € ä�4×#Ñ#×,Ñ,Ó-Ð-r   c                ó    <€ V ^8„  d   QhRS[ /# r    rf   )r   r   s   "€r   r   rŽ   Ñ   s   ø€ ÷ 8ñ 8¡ñ 8r   c                óT   € \        V P                  P                  P                  4      # r@   )rg   r   rº   Ú
z_channelsrÓ   s   &r   Úlatent_channelsÚAudioVAE.latent_channelsÐ   s   € ä�4×#Ñ#×+Ñ+×6Ñ6Ó7Ð7r   c                ó    <€ V ^8„  d   QhRS[ /# r    rf   )r   r   s   "€r   r   rŽ   Õ   s   ø€ ÷ >ñ >¡sñ >r   c                óB   € \        V P                  \        ,          4      # r@   )rg   rc   r—   rÓ   s   &r   Úlatent_frequency_binsÚAudioVAE.latent_frequency_binsÔ   s   € ä�4—=‘=Ô$<Õ<Ó=Ð=r   c                ó    <€ V ^8„  d   QhRS[ /# r    )rÄ   )r   r   s   "€r   r   rŽ   Ù   s   ø€ ÷ Qñ Q¡Eñ Qr   c                óR   € V P                   V P                  ,          \        ,          # r@   )rx   rd   r—   rÓ   s   &r   rÇ   ÚAudioVAE.latents_per_secondØ   s   € à×Ñ $×"5Ñ"5Õ5Ô8PÕPÐPr   c                ó    <€ V ^8„  d   QhRS[ /# r    rf   )r   r   s   "€r   r   rŽ   Ý   s   ø€ ÷ 	Mñ 	M¡Cñ 	Mr   c                óô   € \        V P                  R R4      pVe   \        V4      # \        V P                  RR4      pVf   \        R4      h\        V P                  V,          V P
                  ,          4      # )Úoutput_sample_rateNÚupsample_factorzCVocoder is missing upsample_factor; cannot infer output sample rate)Úgetattrr   rg   ÚAttributeErrorrx   rd   )rD   Úoutput_raterê   s   &  r   ré   ÚAudioVAE.output_sample_rateÜ   sr   € ä˜dŸl™lÐ,@À$ÓGˆØÒ"Ü�{Ó#Ð#Ü! $§,¡,Ð0AÀ4ÓHˆØÒ"Ü ØUóð ô �4×#Ñ# oÕ5¸×8KÑ8KÕKÓLÐLr   c                óJ   € V P                   P                  P                  4       # r@   )Údevice_managerÚpatcherÚ
model_size)rD   Úinput_shapes   &&r   Úmemory_requiredÚAudioVAE.memory_requiredè   s   € Ø×"Ñ"×*Ñ*×5Ñ5Ó7Ð7r   )r   r™   rš   r   )iD¬  )r,   r-   r.   r/   r0   rE   r§   r³   r²   rÈ   r´   Úpropertyrx   rd   rc   rÞ   râ   rÇ   ré   rô   r3   r4   Ú__classcell__)r�   r   s   @@r   rŒ   rŒ   c   sÛ   ù‡ € ÙO÷
ó 
÷<+ò +÷4
ð 
ò

÷Tð T÷	+ð 	+ð ÷3ó ð3ð ÷4ó ð4ð ÷.ó ð.ð ÷8ó ð8ð ÷>ó ð>ð ÷Qó ðQð ÷	Mó ð	M÷8ò 8r   rŒ   )r    Údataclassesr   r;   Úcomfy.audiorq   Ú#comfy.ldm.mmaudio.vae.distributionsr   Ú)comfy.ldm.lightricks.symmetric_patchifierr   Ú1comfy.ldm.lightricks.vae.causal_audio_autoencoderr   r   Ú%comfy.ldm.lightricks.vocoders.vocoderr   r	   r—   r   r7   r`   r<   r=   rŒ   r+   r   r   Ú<module>rþ      sv   ðÛ Ý !Û Û å LÝ E÷÷ JàÐ ñ �$Ô÷Eð Eó ðE÷0Vñ V÷, 4ñ  4ôFF8ˆu�x‰x�‰ö F8r   