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.
arakkoa
Characteristic dimensions (auto, from its own voice profile): ▲ S_STRY ▲ GEND ▲ S_MONO ▲ S_DRAM ▼ S_TECH ▼ REGS ▼ S_ASMR ▼ S_WHIS · emo: Concentration Teasing Malevolence_Malice
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 121/605.
Best 5 of the kept top-20%
R 5.81 · genu 0.49 · blend 0.76 · qual 0.33 · WER 0.76
R 4.99 · genu 0.47 · blend 0.07 · qual 0.46 · WER 0.38
R 4.95 · genu 0.17 · blend 0.58 · qual 0.50 · WER 0.05
R 4.66 · genu 0.09 · blend 0.39 · qual 0.66 · WER 0.11
R 4.43 · genu 0.23 · blend 0.08 · qual 0.59 · WER 0.00
Weakest 5 still inside the top-20% (the selection boundary)
R 1.34 · genu 0.13 · blend 0.38 · qual 0.40 · WER 0.00
R 1.34 · genu 0.14 · blend 0.23 · qual 0.39 · WER 0.06
R 1.32 · genu 0.05 · blend 0.12 · qual 0.55 · WER 0.00
R 1.32 · genu 0.14 · blend 0.36 · qual 0.40 · WER 0.12
R 1.31 · genu 0.11 · blend 0.17 · qual 0.61 · WER 0.03
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.”