Profile-reward recipe: generate 640 (batched) → ASR-filter+trim → SIDON → keep the top 20% by a per-character quality+profile reward → warm-start FT 3 epochs. Below: the quality range of the selected training set, then example generations after each epoch.
phoenix
Characteristic dimensions (auto, from its own voice profile): ▲ GEND ▲ S_STRY ▲ VULN ▲ S_MONO ▼ EXPL ▼ DARC ▼ S_TECH ▼ R_HEAD · emo: Concentration Fatigue_Exhaustion Fear
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 8/28.
Best 5 of the kept top-20%
R 3.03 · genu 0.56 · blend 0.07 · qual 0.52 · WER 0.75
R 2.53 · genu 0.11 · blend 0.50 · qual 0.65 · WER 0.00
R 2.16 · genu 0.20 · blend 0.72 · qual 0.37 · WER 0.37
R 2.02 · genu 0.16 · blend 0.48 · qual 0.49 · WER 0.04
R 1.83 · genu 0.32 · blend 0.00 · qual 0.50 · WER 0.03
Weakest 5 still inside the top-20% (the selection boundary)
R 2.02 · genu 0.16 · blend 0.48 · qual 0.49 · WER 0.04
R 1.83 · genu 0.32 · blend 0.00 · qual 0.50 · WER 0.03
R 1.21 · genu 0.17 · blend 0.21 · qual 0.52 · WER 0.03
R 0.96 · genu 0.21 · blend 0.70 · qual 0.41 · WER 0.69
R 0.95 · genu 0.19 · blend 0.11 · qual 0.54 · WER 0.08
Fantasy line — refined ep1 / ep2 / ep3
“The old road winds past the ruined tower where the last dragon knights once kept their watch.”
Sci-fi line — refined ep1 / ep2 / ep3
“The station drifted in low orbit while the reactor cooled and the scanners swept the empty dark.”
Neutral line — refined ep1 / ep2 / ep3
“The meeting is scheduled for three o'clock and the report is on the desk by the printer.”