Real noise.
Paired truth.
Listen.

The noisy recordings are real aligned inputs; no synthetic corruption or target leakage is added. Super Sonique is the 1,030-hour trajectory-supervised residual-EDM model evaluated with its EMA weights and sealed best sampler: eight Heun intervals, 15 network calls, σ_data 0.5, σ 0.002→2.0, ρ 7, seed 20260831. Its frozen posterior-mean KVAE decoder uses 250-frame cores with 11-frame halos. The comparison is the official single-speaker Sidon v0.1 model—not DialogueSidon—with its own 0.9-peak, 50 Hz high-pass and W2V-BERT preprocessing. Microsoft DNSMOS P.835 OVRL scores the exact MP3s. Recovery is unclamped; n/a means the clean/noisy MOS gap is under 0.1.