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.
seasoned-merchant
Characteristic dimensions (auto, from its own voice profile): ▲ S_MONO ▲ S_NARR ▲ S_STRY ▲ ESTH ▼ EXPL ▼ S_RANT ▼ DARC ▼ DFLU · emo: Concentration Emotional_Numbness Interest
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 127/638.
Best 5 of the kept top-20%
R 6.70 · genu 0.24 · blend 0.50 · qual 0.62 · WER 0.00
R 6.38 · genu 0.16 · blend 0.38 · qual 0.71 · WER 0.00
R 5.16 · genu 0.21 · blend 0.24 · qual 0.67 · WER 0.00
R 4.93 · genu 0.12 · blend 0.35 · qual 0.78 · WER 0.08
R 4.80 · genu 0.17 · blend 0.37 · qual 0.63 · WER 0.00
Weakest 5 still inside the top-20% (the selection boundary)
R 1.52 · genu 0.02 · blend 0.10 · qual 0.77 · WER 0.00
R 1.51 · genu 0.07 · blend 0.01 · qual 0.76 · WER 0.03
R 1.49 · genu 0.01 · blend 0.23 · qual 0.61 · WER 0.00
R 1.49 · genu 0.08 · blend 0.07 · qual 0.69 · WER 0.00
R 1.49 · genu 0.10 · blend 0.07 · qual 0.61 · 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.”