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.
ooze
Characteristic dimensions (auto, from its own voice profile): ▲ S_STRY ▲ S_DRAM ▲ GEND ▲ EMPH ▼ EXPL ▼ S_TECH ▼ S_ASMR ▼ S_WHIS · emo: Interest Concentration Awe
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 121/608.
Best 5 of the kept top-20%
R 5.38 · genu 0.25 · blend 0.81 · qual 0.35 · WER 0.00
R 4.86 · genu 0.19 · blend 0.72 · qual 0.45 · WER 0.00
R 4.73 · genu 0.12 · blend 0.65 · qual 0.54 · WER 0.00
R 4.48 · genu 0.24 · blend 0.54 · qual 0.31 · WER 0.00
R 4.42 · genu 0.26 · blend 0.44 · qual 0.35 · WER 0.00
Weakest 5 still inside the top-20% (the selection boundary)
R 1.44 · genu 0.24 · blend 0.19 · qual 0.30 · WER 0.03
R 1.44 · genu 0.17 · blend 0.04 · qual 0.44 · WER 0.00
R 1.41 · genu 0.18 · blend 0.00 · qual 0.48 · WER 0.00
R 1.40 · genu 0.11 · blend 0.13 · qual 0.49 · WER 0.03
R 1.39 · genu 0.10 · blend 0.11 · qual 0.55 · 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.”