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.
faerie
Characteristic dimensions (auto, from its own voice profile): ▲ S_STRY ▲ REGS ▲ RANG ▲ S_DRAM ▼ S_RANT ▼ S_NEWS ▼ EXPL ▼ GEND · emo: Concentration Hope_Enthusiasm_Optimism Interest
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 124/621.
Best 5 of the kept top-20%
R 4.53 · genu 0.23 · blend 0.25 · qual 0.62 · WER 0.00
R 4.51 · genu 0.22 · blend 0.31 · qual 0.66 · WER 0.00
R 4.43 · genu 0.29 · blend 0.51 · qual 0.52 · WER 0.00
R 4.40 · genu 0.26 · blend 0.31 · qual 0.65 · WER 0.08
R 4.34 · genu 0.20 · blend 0.19 · qual 0.65 · WER 0.00
Weakest 5 still inside the top-20% (the selection boundary)
R 1.22 · genu 0.08 · blend 0.10 · qual 0.63 · WER 0.03
R 1.19 · genu 0.24 · blend 0.17 · qual 0.50 · WER 0.00
R 1.17 · genu 0.16 · blend 0.11 · qual 0.53 · WER 0.04
R 1.16 · genu 0.12 · blend 0.14 · qual 0.59 · WER 0.03
R 1.16 · genu 0.12 · blend 0.11 · qual 0.56 · 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.”