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.
voice-11
Characteristic dimensions (auto, from its own voice profile): ▲ S_STRY ▲ S_MONO ▲ GEND ▲ S_DRAM ▼ REGS ▼ S_CONV ▼ S_ASMR ▼ S_TECH · emo: Concentration Pride Malevolence_Malice
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 126/634.
Best 5 of the kept top-20%
R 6.34 · genu 0.16 · blend 0.59 · qual 0.46 · WER 0.00
R 5.05 · genu 0.14 · blend 0.62 · qual 0.45 · WER 0.00
R 4.78 · genu 0.11 · blend 0.56 · qual 0.54 · WER 0.08
R 4.50 · genu 0.11 · blend 0.42 · qual 0.52 · WER 0.00
R 4.37 · genu 0.16 · blend 0.34 · qual 0.48 · WER 0.00
Weakest 5 still inside the top-20% (the selection boundary)
R 1.22 · genu 0.08 · blend 0.09 · qual 0.48 · WER 0.03
R 1.22 · genu 0.08 · blend 0.25 · qual 0.48 · WER 0.00
R 1.20 · genu 0.10 · blend 0.16 · qual 0.49 · WER 0.00
R 1.19 · genu 0.04 · blend 0.18 · qual 0.60 · WER 0.07
R 1.19 · genu 0.07 · blend 0.04 · qual 0.53 · WER 0.00
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.”