Pj0 |dZddlmZddlZddlZddlmZddlZddl m Z ejjde eejj j eejj j dz ZdZdd d d d d ddddd ZddZedk(r eey)uGenerate the 5 mood reference clips Pocket TTS clones for [emotion] tags. Jarvis can prefix a sentence with [calm]/[warm]/[excited]/[sad]/[dry]/… and have it spoken in that tone. On the Mac (Pocket TTS) that works by *swapping the voice-clone reference clip* per mood — but those clips have to exist. This script renders them once with Chatterbox (which has a real emotion dial), cloning the base Jarvis voice (voices/jarvis_ref.wav) at each mood's exaggeration level, and writes voices/jarvis_.wav. Pocket picks them up automatically at startup. WHERE TO RUN: the RTX/Windows box (CUDA) or this M4 (MPS) — anywhere Chatterbox loads. It is NOT needed for tone *sensing* (that's pure-numpy, always on); it only gives Jarvis's own spoken replies real mood on the Mac Pocket engine. VOICE_ENGINE=chatterbox python tools/make_mood_clips.py # make missing VOICE_ENGINE=chatterbox FORCE=1 python tools/make_mood_clips.py # redo all Stems mirror voice/tts_pocket.py:_TAG_TO_STEM. Chatterbox outputs float32 @ 24kHz. ) annotationsN)Path)wavfilevoicesi]zI[calm] Everything is under control, sir. There's no need to worry at all.zB[warm] Of course, sir. It's always a genuine pleasure to help you.uF[excited] Brilliant news, sir — this is exactly what we'd hoped for!zC[sad] I'm terribly sorry, sir. I do wish I had better news for you.z9[dry] Oh, marvelous. Another meeting. How thrilling, sir.uK[shocked] Good God, sir — you actually did it. I did not see that coming.uT[curious] Now that is genuinely interesting. Tell me more, sir — how does it work?zF[amused] Ha. Oh, that's brilliant, sir. You really are something else.z=[annoyed] Sir. For the third time. It is not going to change.z;[tender] It's alright, sir. Take your time. I'm right here.) calmwarmexcitedsaddryshockedcuriousamusedannoyedtenderc tjjddddlm}m}|s t dytdz }|jst d|d ytjd tjjd d k(}|}d}tjD]*\}}td|dz }|jr|st d|jd@|j|} | t| dk(rt d|jd{t!j"t!j$| t j&dd} t)j*t-|t.| dzj1t j2|dz }t d|jdt| t.z dd-t d|dtdy)N VOICE_ENGINE chatterboxr) ChatterboxTTSloadablezwChatterbox can't load here (no CUDA/MPS headroom). Run this on the RTX box or an idle Mac with VOICE_ENGINE=chatterbox.zjarvis_ref.wavzMissing base voice u — clone Jarvis first.T)exist_okFORCE1jarvis_z.wavz keep z (exists; FORCE=1 to redo)z FAIL z (empty synth))dtypegg?iz wrote z (z.1fzs)u Done — z clip(s) written to z3. Restart Jarvis; Pocket TTS loads them at startup.)osenviron setdefaultvoice.tts_chatterboxrrprint_VOICESis_filemkdirget_MOODSitemsnamesynthlennpclipasarrayfloat32rwritestr_SRastypeint16) rrrefforcettsmadestemlineoutpcms tools/make_mood_clips.pymainr<4sJJ.,7< : I J $ $C ;;= #C5(@AB MM4M JJNN7 #s *E /C Dlln d'$t,, ;;= HSXXJ&@A B iio ;#c(a- HSXXJn5 6 ggbjjBJJ7sC c#hcEk%9%9"((%CD    #c#hsl3%7r:;% Kv1';> >? __main__)returnint)__doc__ __future__rrsyspathlibrnumpyr*scipy.iorpathinsertr/__file__resolveparentr!r0r%r<__name__ SystemExitr=r;rOs&# 3tH~--/66==>? x. " ) ) 0 08 ;  \TXUK]fXOM ! H z TV r=