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.
catfolk
Characteristic dimensions (auto, from its own voice profile): ▲ S_STRY ▲ GEND ▲ S_DRAM ▲ EMPH ▼ S_TECH ▼ S_ASMR ▼ S_WHIS ▼ S_RANT · emo: Concentration Interest Hope_Enthusiasm_Optimism
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 127/637.
Best 5 of the kept top-20%
R 13.02 · genu 0.48 · blend 0.55 · qual 0.56 · WER 0.00
R 6.99 · genu 0.08 · blend 0.61 · qual 0.65 · WER 0.00
R 5.78 · genu 0.14 · blend 0.75 · qual 0.52 · WER 0.00
R 4.96 · genu 0.22 · blend 0.19 · qual 0.57 · WER 0.03
R 4.70 · genu 0.20 · blend 0.26 · qual 0.50 · WER 0.00
Weakest 5 still inside the top-20% (the selection boundary)
R 1.38 · genu 0.11 · blend 0.22 · qual 0.50 · WER 0.07
R 1.36 · genu 0.02 · blend 0.12 · qual 0.58 · WER 0.00
R 1.34 · genu 0.00 · blend 0.27 · qual 0.49 · WER 0.00
R 1.33 · genu 0.12 · blend 0.07 · qual 0.58 · WER 0.00
R 1.32 · genu 0.10 · blend 0.25 · qual 0.53 · WER 0.04
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.”