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.
eradar
Characteristic dimensions (auto, from its own voice profile): ▲ S_STRY ▲ EMPH ▲ S_DRAM ▲ STNC ▼ S_TECH ▼ S_ASMR ▼ S_WHIS ▼ S_NEWS · 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 127/636.
Best 5 of the kept top-20%
R 7.19 · genu 0.21 · blend 0.38 · qual 0.58 · WER 0.00
R 6.07 · genu 0.17 · blend 0.35 · qual 0.59 · WER 0.00
R 5.90 · genu 0.20 · blend 0.35 · qual 0.56 · WER 0.00
R 5.43 · genu 0.14 · blend 0.34 · qual 0.66 · WER 0.00
R 5.30 · genu 0.18 · blend 0.21 · qual 0.68 · WER 0.03
Weakest 5 still inside the top-20% (the selection boundary)
R 1.60 · genu 0.10 · blend 0.00 · qual 0.61 · WER 0.00
R 1.58 · genu 0.08 · blend 0.03 · qual 0.58 · WER 0.00
R 1.58 · genu 0.13 · blend 0.16 · qual 0.43 · WER 0.00
R 1.57 · genu 0.03 · blend 0.07 · qual 0.61 · WER 0.00
R 1.54 · genu 0.12 · blend 0.00 · qual 0.52 · 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.”