nx_phoneme_synth_gate.nx

nx_phoneme_synth_gate.nx -- REFEREE for VOICE-CLONE-001 rung 5a (articulation). Renders a 3-phoneme "word" /a/ -> /i/ -> /u/ and proves the engine ARTICULATES (says a sequence) rather than sustaining one sound: ARTICULATION : the 3 segments must be SPECTRALLY DISTINCT (large pairwise voiceprint distance) -> the voice is moving through sounds. NEG-CONTROL : a constant "word" /a/ /a/ /a/ must have the 3 segments NEARLY IDENTICAL (small distance) -> the variation in the real word comes from the PHONEMES, not from synthesis noise. DETERMINISM : the same sequence renders byte-equal (segment dist == 0). VOICED/UNVOICED : a mixed word /a/(voiced) /s/(unvoiced) /u/(voiced) must read voiced, UNVOICED, voiced (vp frame-count, via nx_pitch). SPEAKER : a second speaker's phoneme inventory (child formants) reading the SAME word differs from speaker 1 AND still articulates. Every value PRINTED. Evidence -> stdout + knowledge/status/phoneme_synth_gate.log. Exit 0 GREEN / 1 RED. Sovereign x86_64 throughout. license_tier: ORIGINAL