nx_voice_read_gate.nx

nx_voice_read_gate.nx -- REFEREE for VOICE-CLONE-001 rung 5-FUSE (voice reads words). Two parametric speakers read two words from a shared phoneme inventory: spk1 : warp 1.00, F0 120 spk2 : warp 1.20, F0 175 WORD_A = /a/ /i/ /u/ WORD_B = /u/ /i/ /a/ (same sounds, reordered) PROVES (measured via rung-1 voiceprint distance, printed): ARTICULATION : each speaker's WORD_A has 3 spectrally-distinct SEGMENTS (>1M) -> the voice reads the sequence, doesn't sustain a sound. SPEAKER IDENTITY DOMINATES CONTENT (the fuse) -- measured on the WHOLE utterance (averaging out per-phoneme context, as real speaker-ID does): same_voice = dist(spkX wordA, spkX wordB) must be SMALL diff_voice = dist(spk1 wordY, spk2 wordY) must be LARGE assert max(same_voice) * 2 < min(diff_voice) i.e. a speaker stays recognizably THEMSELVES across different words, while a different speaker reading the same word is clearly a different voice -> "this voice reads this text." Evidence -> stdout + knowledge/status/voice_read_gate.log. Exit 0 GREEN / 1 RED. Sovereign x86_64 throughout. license_tier: ORIGINAL