+j vddlZddlZddlmZddlmZmZmZddlm Z m Z m Z m Z m Z mZmZmZmZmZmZddlmZmZmZmZmZmZmZmZdZdZdZgd Z d e DZ!d e DZ"Gd d ej#Z$Gddej#Z%Gddej#Z&Gddej#Z'Gddej#Z(Gddej#Z)ej*dddddej+j,dej*dd d!ddej+j,d"ej-d#d$d%&ej*d'd!d!d(dej+j,d)gZ.Gd*d+ej#Z/d,e0d-e1fd.Z2Gd/d0ej#Z3Gd1d2eZ4d-e4fd3Z5dS)4N)override)IOComfyExtensionInput) AddVoiceRequestAddVoiceResponse DialogueInputDialogueSettingsSpeechToSpeechRequestSpeechToTextRequestSpeechToTextResponseTextToDialogueRequestTextToSoundEffectsRequestTextToSpeechRequestTextToSpeechVoiceSettings) ApiEndpointaudio_bytes_to_audio_inputaudio_ndarray_to_bytesio"audio_tensor_to_contiguous_ndarraysync_op sync_op_rawupload_audio_to_comfyapivalidate_stringELEVENLABS_MUSIC_SECTIONSELEVENLABS_COMPOSITION_PLANELEVENLABS_VOICE))CwhRBWXzGAHq8TQ4Fs17Rogermaleamerican)EXAVITQu4vr4xnSDxMaLSarahfemaler )FGY2WhTYpPnrIDTdsKH5Laurar#r )IKne3meq5aSn9XLyUdCDCharlier australian)JBFqnCBsd6RMkjVDRZzbGeorgerbritish)N2lVS1w4EtoT3dr4eOWOCallumrr )SAz9YHcvj6GT2YYXdXwwRiverneutralr )SOYHLrjzK2X1ezoPC6crHarryrr )TX3LPaxmHKxFdv7VOQHJLiamrr )Xb7hH8MSUJpSbSDYk0k2Alicer#r+)XrExE9yKIg1WjnnlVkGXMatildar#r )bIHbv24MWmeRgasZH58oWillrr )cgSgspJ2msm6clMCkdW9Jessicar#r )cjVigY5qzO86Huf0OWalEricrr )hpp4J3VqNfWAUOO0d1UsBellar#r )iP95p4xoKVk53GoZ742BChrisrr )nPczCjzI2devNBz1zQrbBrianrr )onwK4e9ZLuTAKqWW03F9Danielrr+)pFZP5JQG7iQjIQuC4BkuLilyr#r+)pNInz6obpgDQGcFmaJgBAdamrr )pqHfZKP75CvOlQylNhV4Billrr c.g|]\}}}}|d|d|dSz (z, )).0_namegenderaccents @/home/wildlama/comfy/ComfyUI/comfy_api_nodes/nodes_elevenlabs.py rW<s;lll?Vq$PVt::v:::::lllc0i|]\}}}}|d|d|d|SrNrP)rQvoice_idrSrTrUs rV r[=sH1O4t""v"""""HrXc xeZdZedejfdZedejde de de de dej f dZ d S) ElevenLabsSpeechToTextreturnctjddddtjddtjdtjd tjd d d tjdd d tjdddddtjj dtjdddddtjj dtj dgdddggd !tj d"d#d$ tj d%d&d&d'tjj d()tj d*d+d&d,d-.gtj d/0tj d"0tj d10gtjjtjjtjjgd2tjd345 S)6Nr]zElevenLabs Speech to Textpartner/audio/ElevenLabsznTranscribe audio to text. Supports automatic language detection, speaker diarization, and audio event tagging.audiozAudio to transcribe.tooltipmodel scribe_v2tag_audio_eventsFz=Annotate sounds like (laughter), (music), etc. in transcript.defaultrcdiarizez"Annotate which speaker is talking.diarization_thresholdg)\(?皙?g?{Gz?zSSpeaker separation sensitivity. Lower values are more sensitive to speaker changes.rhminmaxstep display_moderc temperatureg@zNRandomness control. 0.0 uses model default. Higher values increase randomness.timestamps_granularity)word characternoneruz&Timing precision for transcript words.)optionsrhrczModel to use for transcription.rxrc language_codedISO-639-1 or ISO-639-3 language code (e.g., 'en', 'es', 'fra'). Leave empty for automatic detection. num_speakersr zHMaximum number of speakers to predict. Set to 0 for automatic detection.)rhrnrorqrcseed6Seed for reproducibility (determinism not guaranteed).rhrnrorctext display_name words_jsonTzL{"type":"usd","usd":0.0073,"format":{"approximate":true,"suffix":"/minute"}}expr node_idrcategory descriptioninputsoutputshidden is_api_node price_badge)rSchemaAudior DynamicComboOptionBooleanFloat NumberDisplaysliderComboStringIntOutputHiddenauth_token_comfy_orgapi_key_comfy_org unique_id PriceBadgeclss rV define_schemaz$ElevenLabsSpeechToText.define_schemaCsy,4/c2%%..' " 0 0$6,1,k!1!"!" !# 0 0$-,1,P!1!"!" !#$;,0(+(+)-131A1H-Z!/ !" !"!#$1,/(+(+)-131A1H-a!/ !" !"!#$<,I,I,I,2,T !/!"!"?%((*V>[&..^ #;   "!#!1!8f  "T EIV   f 55   o >>   l ;;  . + #  gw^ ^ ^ ^ rXrardrzr}rc *K|dr|rtdt|dt||dddd{V|r|nd|d|d kr|nd|d |d|dr|d nd||d  }t |t ddt |dd{V}tj|j rd|j Dngd}tj |j |j |S)NrizxNumber of speakers cannot be specified when diarization is enabled. Either disable diarization or set num_speakers to 0.rdmp4aac audio/mp4)container_format codec_name mime_typerfrrtrjrr) model_idcloud_storage_urlrzrfr}rtrirjrrrz#/proxy/elevenlabs/v1/speech-to-textPOSTpathmethodmultipart/form-data)response_modeldata content_typec:g|]}|dS)T exclude_none) model_dump)rQws rVrWz2ElevenLabsSpeechToText.execute..s& E E EQ\\t\ , , E E ErX)indent) ValueErrorr rstriprrr jsondumpswordsr NodeOutputrrz) rrardrzr}rrequestresponsers rVexecutezElevenLabsSpeechToText.executes     G &7^$<UUuP[%%%,9+>+>+@+@J--d"#56)5)9)9t#()A#B)$DI)DT"^%(?"@"@Z^m,    !  B6 R R R/.          ZIQ _ E Ehn E E E E]_   }X]H,BJOOOrXN)__name__ __module__ __qualname__ classmethodrrrrrdictstrintrrrPrXrVr]r]Bs_ bi_ _ _ [_ B&P{&P&P &P  &P  &P &P&P&P[&P&P&PrXr]c^eZdZedejfdZededejfdZ dS)ElevenLabsVoiceSelectorr^c tjddddtjdtdgtjt dgd S) NrzElevenLabs Voice Selectorr`zCSelect a predefined ElevenLabs voice for text-to-speech generation.voicez5Choose a voice from the predefined ElevenLabs voices.ryrF)rrrrrrr)rrrrELEVENLABS_VOICE_OPTIONSCustomrrrs rVrz%ElevenLabsVoiceSelector.define_schemasxy-4/]4S *++222HH    rXrct|}|std|tj|S)NzUnknown voice: )ELEVENLABS_VOICE_MAPgetrrr)rrrZs rVrzElevenLabsVoiceSelector.executesD'++E22 86u6677 7}X&&&rXN) rrrrrrrrrrrPrXrVrrsk bi   [ &'C'BM'''['''rXrczeZdZedejfdZededededede ded e d edej fd Z d S) ElevenLabsTextToSpeechr^ctjddddtjtddtjdd d d tjd ddddtjjdtj dgddtj dtj dtjdddddtjjdtjdddddtjjd tj d!d"d#$tjd%ddd&dtjjd'gtj d(tjdddddtjjdtjdddddtjjd ggd)tjd*d d+$tj d,d-d.d/d01tj d2d3d4gd5gtjgtjjtjjtjjgd tjd678 S)9NrzElevenLabs Text to Speechr`zConvert text to speech.rzVVoice to use for speech synthesis. Connect from Voice Selector or Instant Voice Clone.rbrTr{zThe text to convert to speech. multilinerhrc stability?rs?rlVoice stability. Lower values give broader emotional range, higher values produce more consistent but potentially monotonous speech.rmapply_text_normalizationautoonoffjText normalization mode. 'auto' lets the system decide, 'on' always applies normalization, 'off' skips it.ryrdeleven_multilingual_v2speedffffff??6Speech speed. 1.0 is normal, <1.0 slower, >1.0 faster.similarity_boost?LSimilarity boost. Higher values make the voice more similar to the original.use_speaker_boostF/Boost similarity to the original speaker voice.rgstyle皙?YStyle exaggeration. Higher values increase stylistic expression but may reduce stability. eleven_v3z Model to use for text-to-speech.rzr|rrrrrr output_format mp3_44100_192opus_48000_192Audio output format.L{"type":"usd","usd":0.24,"format":{"approximate":true,"suffix":"/1K chars"}}rr)rrrrrrrrrrrrrrrrrrrrrrs rVrz$ElevenLabsTextToSpeech.define_schemasHy,4/1 *++11t2 "<   !#!1!8_  .111I %%..4 "$+,/(+(+)-131A1H,d!/!"!"!#$6,0(+(+)-131A1H,z!/!"!"!# 0 0$7,1,]!1!"!" !#$+,/(+(+)-131A1H-@!/ !" !"1"%%L..' "$+,/(+(+)-131A1H,d!/!"!"!#$6,0(+(+)-131A1H,z!/!"!"O>~?C&BBF #;   "T #,.>?2Wpd!! . + #  gB B B B rXrrrrrdrzrrc Kt|dt||d|r|ndt||d|d|dd|dd|| } t |t d |d d |i | dd{V} tjt| S)Nr min_lengthrdrrrr)rrrrr)rrrzvoice_settingsrrz$/proxy/elevenlabs/v1/text-to-speech/rrrr query_paramsTr as_binary) rrrrrrrrrr) rrrrrrdrzrrrrs rVrzElevenLabsTextToSpeech.executess# ++++%7^+8+>+>+@+@J--d4#!&'9!:Gn"')),?"F"Fii.. %=    %  CECC-}=              }7AABBBrXN) rrrrrrrrfloatrrrrrPrXrVrrsC biC C C [C J$C$C$C $C #& $C  $C$C$C$C $C$C$C[$C$C$CrXrcheZdZedejfdZedejdej fdZ dS)ElevenLabsAudioIsolationr^c <tjddddtjddgtjgtjjtjjtjjgdtj d  S) NrzElevenLabs Voice Isolationr`z?Remove background noise from audio, isolating vocals or speech.raz.Audio to process for background noise removal.rbTJ{"type":"usd","usd":0.24,"format":{"approximate":true,"suffix":"/minute"}}rr) rrrrrrrrrrrs rVrz&ElevenLabsAudioIsolation.define_schemasy.5/YL!! . + #  e)    rXracKt|d}t||ddd}t|tdddd |d fid d d{V}t jt |S)Nwaveform sample_raterrz$/proxy/elevenlabs/v1/audio-isolationrrra audio.mp4rrT)filesrr)rrrrrrr)rra audio_data_npaudio_bytes_iors rVrz ElevenLabsAudioIsolation.executes ;5;LMM 1-}AUW\^cdd$  CF S S S[.+FG.          }7AABBBrXN) rrrrrrrrrrrrPrXrVrrsy bi   [ 4 C{ C  C C C[ C C CrXrc feZdZedejfdZededededej fdZ dS)ElevenLabsTextToSoundEffectsr^ctjddddtjdddd tjd tjd tjd d dddtjjdtj dddtjdddddtjjdggdtj dd d!gd"gtj gtj jtj jtj jgdtjd#$% S)&Nrz ElevenLabs Text to Sound Effectsr`z.Generate sound effects from text descriptions.rTr{z1Text description of the sound effect to generate.rrd eleven_sfx_v2durationg@rg>@rkz'Duration of generated sound in seconds.rmloopFz'Create a smoothly looping sound effect.rgprompt_influenceg333333?rsrrlzeHow closely generation follows the prompt. Higher values make the sound follow the text more closely.z)Model to use for sound effect generation.ryrrrrzJ{"type":"usd","usd":0.14,"format":{"approximate":true,"suffix":"/minute"}}rr)rrrrrrrrrrrrrrrrrrrs rVrz*ElevenLabsTextToSoundEffects.define_schemasy2;/H "O   %%..+ "$.,/(+(,),131A1H,U!/!"!"!# 0 0$*,1,U!1!"!" !#$6,/(+(+)-131A1H-a!/ !" !">HC&""F#,.>?2U/b!! . + #  e}A A A A rXrrdrc "Kt|dt|tddd|it||d|d|d d d d{V}t jt|S) Nrrz%/proxy/elevenlabs/v1/sound-generationrrrrrr)rduration_secondsrrTr)rrrrrrrr)rrrdrrs rVrz$ElevenLabsTextToSoundEffects.executes ++++$  <-}=    +!&z!2!&'9!:YYvt,,           }7AABBBrXN) rrrrrrrrrrrrPrXrVrrsB biB B B [B HCCC C  CCC[CCCrXrcveZdZedejfdZedejjde dej fdZ dS)ElevenLabsInstantVoiceCloner^ctjddddtjdtjtjdddd d tjd d dgtjt dgtj j tj j tj j gdtjd S)NrzElevenLabs Instant Voice Cloner`z]Create a cloned voice from audio samples. Provide 1-8 audio recordings of the voice to clone.r rar)prefixrnroz#Audio recordings for voice cloning.)templatercremove_background_noiseFzARemove background noise from voice samples using audio isolation.rgrrTz{"type":"usd","usd":0.15}rr)rrAutogrowrTemplatePrefixrrrrrrrrrrrs rVrz)ElevenLabsInstantVoiceClone.define_schema)s y19/B !![77w//& 8 B"     -!_!$ *++222HH . + #  +JKKKA! ! ! ! rXr rc Kg}|D]j}||}|d}|d}t|}t||dd} |d|d| dffkt |t dd t tttj | |d d{V} tj | j S)Nr rrrr z.mp4rz/proxy/elevenlabs/v1/voices/addrr)rSrr)rrr r)rrappendgetvaluerrrrruuiduuid4rrrZ) rr r file_tupleskeyrar rr r rs rVrz#ElevenLabsInstantVoiceClone.executeNs! AC  b bC#JE$]3KZ(H>xHHM5m[RWY^__N   S,,,8O8O8Q8QS^)_` a a a a  >v N N N+ &&(?.          }X.///rXN) rrrrrrrrTypeboolrrrPrXrVrr(s" bi" " " [" H0{0"&0  000[000rXrrrrrrlrrmrrrsrrFrrgrrrceZdZedejfdZededej de de dede d e dejfd Zd S) ElevenLabsSpeechToSpeechr^ctjddddtjtddtjdd tjd d d d dtjjdtj dtj dttj dtgdtj dddgdtj dddddtjdd d!"gtjgtjjtjjtjjgd#tjd$%& S)'Nr*zElevenLabs Speech to Speechr`z]Transform speech from one voice to another while preserving the original content and emotion.rzXTarget voice for the transformation. Connect from Voice Selector or Instant Voice Clone.rbrazSource audio to transform.rrrsrrlrrmrdeleven_multilingual_sts_v2eleven_english_sts_v2z1Model to use for speech-to-speech transformation.ryrrrrrrSeed for reproducibility.rrFz?Remove background noise from input audio using audio isolation.rgTrrr)rrrrrrrrrrrELEVENLABS_STS_VOICE_SETTINGSrrrrrrrrrrs rVrz&ElevenLabsSpeechToSpeech.define_schemasy.6/w *++11J2 8!#!1!8_  %%..89..39 P&  #,.>?2  "7    -!]!]3j!! . + #  eEE E E E rXrrarrdrrrc Kt|d}t||ddd} t||d|d|d|d } t|t d |d d |i t |d| d||dd| dfiddd{V} tj t| S)Nrr rrrrrr)rrrrrz&/proxy/elevenlabs/v1/speech-to-speech/rrrrdTr)rrrrrar rr)rr rr) rrrrrr model_dump_jsonr"rrr) rrrarrdrrrr r rrs rVrz ElevenLabsSpeechToSpeech.executes=;5;LMM 1-}AUW\^cdd2"#56.#$78.    %  EeEE-}=    'w-==4=PP(?  [.*A*A*C*C[QR.         "}7AABBBrXN)rrrrrrrrrrrrrr(rrrPrXrVr*r*sF biF F F [F P$C$C{$C $C  $C  $C$C"&$C $C$C$C[$C$C$CrXr*countr^c g}td|dzD]t}|tjd|ddd|dtjt d|d |d  gu|S) z>Generate input widgets for a given number of dialogue entries.rrTr{z Text content for dialogue entry .rrzVoice for dialogue entry z5. Connect from Voice Selector or Instant Voice Clone.rb)rangeextendrrrrr)r3ris rV_generate_dialogue_inputsr9s F 1eai     1JJ"CqCCC    *++11AKKpppp2   MrXcveZdZedejfdZedededede dede d edej fd Z d S) ElevenLabsTextToDialoguer^cntjddddtjdddddtjjd tjd gd d tjddgdtjdtjdtdtjdtdtjdtdtjdtdtjdtdtjdtdtjdtd tjd!td"tjd#td$tjd%td&g d'tj d(d)d*+tj d,dd-d.d/0tjd1d2d3gd4gtj gtjjtjjtjjgd5tjd678 S)9Nr;zElevenLabs Text to Dialoguer`zZGenerate multi-speaker dialogue from text. Each dialogue entry has its own text and voice.rrrsrrrmrrrryrdrz%Model to use for dialogue generation.r1r2r345678r9 10 zNumber of dialogue entries.rzr{r|rgrrr.r/rrrrrTrrr)rrrrrrrrrr9rrrrrrrrrrs rVrz&ElevenLabsTextToDialogue.define_schemasy.6/t!#!1!8_  .111I (MC %%..s4Ma4P4PQQ..s4Ma4P4PQQ..s4Ma4P4PQQ..s4Ma4P4PQQ..s4Ma4P4PQQ..s4Ma4P4PQQ..s4Ma4P4PQQ..s4Ma4P4PQQ..s4Ma4P4PQQ..t5Nr5R5RSS :&  #;   "7 #,.>?2g8t!! . + #  gOJ J J J rXrrrdrrzrrc Kt|d}g} td|dzD]M} |d| } |d| } t| d| t | | Nt | ||r|ndt|||} t|td d d |i | d d{V}tj t|S)Nrrrrr)rrZ)r)rrrzsettingsrrz%/proxy/elevenlabs/v1/text-to-dialoguerrrTr) rr6rr!r rrr rrrrr)rrrrdrrzrr num_entriesdialogue_inputsr8rrZrrs rVrz ElevenLabsTextToDialogue.executeesU&*++ /1q+/** P PA***%Dkakk*H DQ / / / /  " "=dX#N#N#N O O O O'"+8+>+>+@+@J--d% :::%=    %  <-}=              }7AABBBrXN) rrrrrrrrrrrrrrPrXrVr;r;sK biK K K [K Z#C#C#&#C #C  #C  #C#C#C #C#C#C[#C#C#CrXr;cLeZdZedeeejfdZdS)ElevenLabsExtensionr^cjKtttttt t tgSN)r]rrrrrr*r;)selfs rV get_node_listz!ElevenLabsExtension.get_node_lists* # # " $ ( ' $ $  rXN) rrrrlisttyper ComfyNoderWrPrXrVrSrSsE   T$r|*<%=    X    rXrSc"KtSrU)rSrPrXrVcomfy_entrypointr\s   rX)6rr#typing_extensionsrcomfy_api.latestrrrcomfy_api_nodes.apis.elevenlabsrrr r r r r rrrrcomfy_api_nodes.utilrrrrrrrrrrrELEVENLABS_VOICESrrrZr]rrrrrrrrrr0r*rrXr9r;rSr\rPrXrVrbs &&&&&&6666666666                                              8;%0mlZklllSd JPJPJPJPJPR\JPJPJPZ'''''bl''':lClClClClCR\lClClC^*C*C*C*C*Cr|*C*C*CZ]C]C]C]C]C2<]C]C]C@@0@0@0@0@0",@0@0@0HHNN   %,HHNN   %,^JA HNN   %,k1!!HoCoCoCoCoCr|oCoCoCdST*sCsCsCsCsCr|sCsCsCl      .    ! 3!!!!!!rX