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.
monkey
Characteristic dimensions (auto, from its own voice profile): ▲ S_STRY ▲ S_DRAM ▲ EMPH ▲ AROU ▼ S_TECH ▼ S_ASMR ▼ S_WHIS ▼ S_NEWS · emo: Amusement Interest Concentration
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 122/613.
Best 5 of the kept top-20%
R 4.76 · genu 0.18 · blend 0.12 · qual 0.57 · WER 0.00
R 4.71 · genu 0.20 · blend 0.25 · qual 0.48 · WER 0.00
R 4.66 · genu 0.18 · blend 0.51 · qual 0.29 · WER 0.04
R 4.48 · genu 0.14 · blend 0.28 · qual 0.52 · WER 0.08
R 4.10 · genu 0.23 · blend 0.51 · qual 0.59 · WER 0.77
Weakest 5 still inside the top-20% (the selection boundary)
R 1.40 · genu 0.09 · blend 0.00 · qual 0.56 · WER 0.00
R 1.37 · genu 0.08 · blend 0.00 · qual 0.53 · WER 0.00
R 1.36 · genu 0.05 · blend 0.05 · qual 0.51 · WER 0.00
R 1.36 · genu 0.06 · blend 0.03 · qual 0.55 · WER 0.03
R 1.35 · genu 0.10 · blend 0.00 · qual 0.60 · WER 0.20
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.”