U ^;j)Y@sdZddlmZddlZddlZddlZddlZddlmZddl m Z e dZ Gddde Zed d Zd d d dgddddgddddgddddgddddgdZd ddddgZd dddddZd dddddZd d!Zd"d#d#d#d#d#d$d$d$d$d$d%d%d&d&d&d'Zed(ZdVd)d)d)d)d*d+d,d-Zd.d/d0d1Zd*d*d2d3d4Zed5d6Zd*d)d7d8d9d:ZdWd*dd?Zd*d)d@d*dAdBdCZ dXd*dDd*d*d*d*dEd*d*dF dGdHZ!dYd*dDd*dJdKdLZ"dZdOdPd*dpsz+dominant_base_for_turns..rr)Counteremotionvadcsg|]\}}|kr|qSr r )rbc) top_countr r r~scs |dS)Nr)r)r) arousal_sumr r z)dominant_base_for_turns..)key) collectionsrrEMOTION_TO_BASElenabsfloatmaxvaluesitems) turnsrealrcountsremobaseraZtiedr )r"r!r dominant_base_for_turnsgs$      r4rrrr r)rrlaughter excitement agreement gratitudercryingsulkapologystruggler refusalrflustershyZELEVENLABS_VOICE_IDz str | Nonestr)roverrideseed pool_overridereturncCs~|r|StrtS|dkrt}n&|r,t|}nt|p6dd}t|}|dkrT|dStt|d d}||t |S)uPick voice ID from pool keyed by emotion → base emotion. seed: stable key (e.g. text or scenario_id+turn_idx) to deterministically pick a voice from the pool. Without seed, picks first voice in pool. Deterministic selection ensures same turn always gets same voice on re-run. pool_override: force voice selection from a specific base-emotion pool (e.g. 'neutral' for monologues — same voice across all emotions in turn). Zfemale_monologuerNrutf-8) SINGLE_VOICE_OVERRIDEFEMALE_MONOLOGUE_POOLVOICE_POOLS_BY_BASEr'rinthashlibmd5encode hexdigestr()rrArBrCpoolr2hr r r voice_id_for_emotions  rQz list | None)rcCsv|dkrdddddS|\}}}tt|}ttdtdd d |}ttd tddd |}d}|||ddS) uMap VAD → ElevenLabs voice_settings for expressive prosody. For v3 + audio tags: tags carry the emotion, voice_settings carry consistency. Stacking high `style` on top of expressive tags triple-amplifies expression and produces distortion (raspy / accent drift / foreigner-sounding) — that was the failure mode observed on 2026-05-07. We keep style modest and stability solid. Intensity is |A| magnitude — both high-positive (excited) and low-negative (sad) arousal scale prosody equally. Ng?g?g333333?T) stabilitysimilarity_booststyleZuse_speaker_boostrg?g?g?g?g?)r)r*r+min)rVAD intensityrTrRrSr r r _build_voice_settingss   rZ)rrDcCs$|}|r |ddkr |dS|S)uFEnsure text ends with sentence-ending punctuation so v3 doesn't cut the tail. Note: `、` (U+3001) is a Japanese reading-comma, not a Korean sentence terminator — it was here by mistake. Korean uses ASCII `.?!` plus sometimes `…`. CJK terminators `。？！` are kept as a safety net for pasted text. u.?!。？！….)r)rr r r _ensure_trailing_punctsr]reu9[ㅠㅜ]{2,}|[ㅋㅎ]{2,}|[ㅏㅓㅗㅡㅣ]{2,}|[ㅡ]{2,}ztuple[str, bool, bool])rrrDcCsnddl}t|d|}t|d|}|dd|}|dd|}|dd|}|dd|}|||fS) uv3 reads bare Korean jamo runs (ㅠㅠ, ㅋㅋ, ㅎㅎ) literally — comes out as "어어/유유/크크" instead of sobbing/laughing. Replace with v3 audio tags so they render as real reactions. ㅠㅠ / ㅜㅜ → [crying] (inline, at the emoticon position. Deleting it and relying on the head [crying]/[sorrowful] tag produced NO audible sob at all, so inline stays. It's imperfect — v3 occasionally rolls it as a laugh or a sigh-with-gap — but at least the cry is present. Re-roll bad takes rather than remove.) ㅋㅋ / ㅎㅎ → stripped (removed entirely, NOT tagged — the clip is still emotion=laughter so the head [laughs] tag carries the laugh; we just don't want the literal ㅋㅋ jamo in the text at all) bare vowel runs (ㅏㅏ, ㅡㅡ…) → stripped (garbage) No leading period — that produced a hard sentence break before the reaction (audible "speak…long pause…laugh" instead of "speak laugh"). The canonical tags fire mid-clause without it. Returns (cleaned_text, had_sob, had_chuckle) — flags reserved but head-tag suppression is OFF by default now. Head tag carries the emotional prosody for the whole line; the body tag just stamps the reaction sound. Suppressing the head was making the voiced emotion drop out entirely. rNu [ㅠㅜ]{2,}u [ㅋㅎ]{2,}z [crying]  u[ㅏㅓㅗㅡㅣ]{2,}z\s+)r^boolsearchsubr)rrr^had_sob had_chuckler r r _scrub_korean_emoticonssre rJ)r min_charsrDcCs"t||kr|ddS|S)uv3 fails / hallucinates on very short inputs (e.g. "응.", "네."). Pad with neutral filler so the model has enough context. We add an ellipsis-style trailer that v3 reads as a brief pause, not extra speech. r\z...)r(rrstrip)rrgr r r _pad_short_textsrir`)rr use_v3_tagsrDcCsddl}|dd|}t||\}}}t|}t|}|rF|rF|dkrJ|Sdddddd d d d d dddddd}||dp|d}|r|dkrd}|rd|d|S|S)uxPrepend emotion tag for v3 models (supports bracketed audio tags). Adds a leading period buffer before the tag — v3 sometimes eats the first token after an audio tag, which truncates the first word of real content. Also ensures trailing punctuation so the model doesn't cut the final word. Short texts are padded — v3 hallucinates on <12-char inputs. rNz \[[^\]]*\]r_rz [happily]z[laughs]z [excited]z [sorrowful]z[crying]z[sighs]z[resigned tone]z [frustrated]z[angry]z[flatly]z [surprised]z [stammers]z [whispers])rr5r6r7r8rr9r:r;r<r r=rr>r?r)rr9r:r;z. )r^rbrerir]rr)rrrjr^rcrdZtag_maptagr r r _format_text_with_emotions6  rlrlist) rout_pathvoice_idmodelapi_keyrr voice_seed voice_poolc sddlmt||pj|dp(tp6tdsDtddk} t ||| t |t } fdd} dd l } d } t| D]:}z| d | Id HWd Stk r}zt|d kp d kp d kp dkp dkp dkp dk}|rzFtdjddt|jdt|dddt|ddd ddWntk rYnXtdd d|dkpd k}d!kpd"kpd#k}tfd$d%d&D}tfd'd%d(D}|p |p |p |}|r|| d)kr|r,d*nd+}td,|d-|| dd.}|rXd!n|rbdn |rld/nd0}td1|d2d d3d4|d)d5| d6|d7d8 t |Id HWYqW5d }~XYqXqd S)9u2Synthesize one clip with ElevenLabs, using emotion+VAD for prosody. If voice_id is None, picks voice automatically based on emotion → base mapping. voice_seed: stable key for deterministic voice pick from the pool. voice_pool: override base-emotion pool (e.g. 'neutral' for monologues). r) ElevenLabs)rArBrCZELEVENLABS_API_KEYz#ELEVENLABS_API_KEY env var requiredZv3csddl}d}|jjdd}d}|d}|D]}||qBW5QRXz@|jddd d d t|d d ddddtg dd|jddWnDt |j fk r}z t d|d| W5d}~XYnXdS)Nr)rqZ mp3_44100_128)rorZmodel_idZ output_formatvoice_settingsz.raw.mp3wbZffmpegz-yz -loglevelerrorz-iz-afz adelay=180|180,apad=pad_dur=0.18z-c:aZ libmp3lamez-b:aZ128kT)check) missing_okz[tts] ffmpeg pad failed (z); saving raw audio) subprocessZtext_to_speechconvert with_suffixopenwriterunr@unlinkFileNotFoundErrorCalledProcessErrorLOGwarningreplace)rzZclientZaudioZraw_pathfchunke)rtrqformatted_textrprnrorur r _syncasF    z#synth_one_elevenlabs.._syncNZ401Z402Z403ZquotaZ unauthorizedZpayment_requiredZbillingz/tmp/tts_quota_error.txtz=== full exception === z === type === z === status_code attr === Z status_codezn/az === body attr === body rE)encodingzElevenLabs auth/quota error: Z409Zalready_runningZ429Z rate_limitZtoo_many_requestsc3s|]}|kVqdSNr rr )msgr r sz'synth_one_elevenlabs..)z 500 z 502 z 503 z 504 c3s|]}|kVqdSrr r)lowr r rs)timeoutZ connectionzremote disconnectedrg @g@gN@rg?Z5xxnet[tts] z on voice u …, retry /z in z.1fs)Zelevenlabs.clientrtrQstemDEFAULT_EL_MODELosgetenv RuntimeErrorlowerrlrZasyncioget_event_looprandomrangerun_in_executor Exceptionr@r write_texttypergetattrranyrUuniformrinfosleep)rrnrorprqrrrrrsrjlooprrZ max_retriesZattemptrZis_quotaZis_409Zis_429Zis_5xxZis_netZ retryabler2delaykindr ) rtrqrrrprrnrorur synth_one_elevenlabsDsl   /2 "8rko-KR-SunHiNeural)rrnvoicec shtjdd|d|dt|tjjtjjd IdH}|IdH\}}|jdkrdtd| dd dS) zFree fallback via edge-tts CLI.zedge-ttsz--voicez--textz --write-media)stdoutstderrNrzedge-tts failed: r) rcreate_subprocess_execr@rzDEVNULLPIPE communicate returncoderdecode)rrnrproc_errr r r synth_one_edges  r elevenlabsz List[str]z List[Path]z List[list]z List[bool]) texts out_pathsbackend concurrencyemotionsvads voice_seeds voice_pools voice_idsrDc st tkstt|dgt dddddkrFtnt dd fdd tjfd d tt DId Hrtd dt ddrtddd ddS)zParallel TTS for a batch. Returns success flag per item. emotions/vads: per-item lists aligned with texts. If provided, passed into the TTS call for emotional prosody. Frr)flagreasonrrJ)ic s4IdHdr,W5QIdHRdS|}|rn|jdkrnd|<d7W5QIdHRdSzt}dkrdk r||d< dk r ||d< dk rʈ |dk rʈ ||d< dk r |dk r ||d < dk r |dk r ||d < ||f|IdHd|<Wntk rx}z(dd<t|d <td |W5d}~XYn:tk r}zt d |d|W5d}~XYnXW5QIdHRXdS)NriTrrrrrrrsroru*[tts] QUOTA EXHAUSTED — aborting batch: z [tts] fail z: ) existsstatst_sizedictrr@rrwrr)rpathZ per_kwargsr)abortedrrkwargsokr semaphoreskippedsynth_fnrrrrrr r ones:         zsynth_all..onecsg|] }|qSr r )rr)rr r rszsynth_all..Nrrz turns skipped (already on disk)ru< [tts] ⚠ batch aborted on quota/auth error. reason: rruz partial results saved. Re-run with a different ELEVENLABS_API_KEY to resume — already-done turns will be skipped.) r(AssertionErrorr Semaphorerrgatherrprint) rrrrrrrrrrr )rrrrrrrrrrrrrrrr synth_alls  ,!&r)NNN)rf)NNNNNNN)r)rrNNNNN)$r __future__rrrKloggingrpathlibrtypingr getLoggerrrrrrrIrHFEMALE_BY_BASE MALE_BY_BASEr4r'rGrQrZr] __import__compileZ_KOREAN_EMOTICON_RErerirlrrrr r r r s     &    % 2 u