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.
gravelly-orc-warlord
Characteristic dimensions (auto, from its own voice profile): ▲ S_STRY ▲ S_MONO ▲ S_NARR ▲ GEND ▼ S_RANT ▼ DARC ▼ EXPL ▼ S_TECH · emo: Concentration Interest Emotional_Numbness
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 115/576.
Best 5 of the kept top-20%
R 5.02 · genu 0.37 · blend 0.27 · qual 0.53 · WER 0.03
R 4.32 · genu 0.16 · blend 0.55 · qual 0.60 · WER 0.00
R 4.28 · genu 0.14 · blend 0.82 · qual 0.47 · WER 0.00
R 4.19 · genu 0.07 · blend 0.60 · qual 0.63 · WER 0.00
R 3.92 · genu 0.09 · blend 0.53 · qual 0.67 · WER 0.00
Weakest 5 still inside the top-20% (the selection boundary)
R 1.45 · genu 0.13 · blend 0.04 · qual 0.73 · WER 0.00
R 1.43 · genu 0.15 · blend 0.61 · qual 0.41 · WER 0.00
R 1.41 · genu 0.18 · blend 0.00 · qual 0.58 · WER 0.00
R 1.40 · genu 0.07 · blend 0.33 · qual 0.51 · WER 0.00
R 1.39 · genu 0.06 · blend 0.15 · qual 0.66 · WER 0.06
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.”