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.
druid
Characteristic dimensions (auto, from its own voice profile): ▲ S_STRY ▲ EMPH ▲ GEND ▲ S_DRAM ▼ S_CONV ▼ S_TECH ▼ S_ASMR ▼ S_WHIS · emo: Pride Concentration Malevolence_Malice
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 125/627.
Best 5 of the kept top-20%
R 6.76 · genu 0.17 · blend 0.43 · qual 0.49 · WER 0.04
R 6.26 · genu 0.20 · blend 0.95 · qual 0.50 · WER 0.79
R 6.08 · genu 0.16 · blend 0.35 · qual 0.60 · WER 0.00
R 5.56 · genu 0.09 · blend 0.35 · qual 0.66 · WER 0.00
R 5.18 · genu 0.14 · blend 0.31 · qual 0.46 · WER 0.00
Weakest 5 still inside the top-20% (the selection boundary)
R 1.29 · genu 0.09 · blend 0.29 · qual 0.38 · WER 0.00
R 1.29 · genu 0.06 · blend 0.00 · qual 0.56 · WER 0.03
R 1.26 · genu 0.07 · blend 0.06 · qual 0.47 · WER 0.07
R 1.25 · genu 0.09 · blend 0.01 · qual 0.50 · WER 0.03
R 1.23 · genu 0.10 · blend 0.16 · qual 0.42 · 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.”