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.
voice-14
Characteristic dimensions (auto, from its own voice profile): ▲ S_NARR ▲ S_MONO ▲ AGEV ▲ S_STRY ▼ DARC ▼ S_RANT ▼ EXPL ▼ S_PLAY · emo: Concentration Emotional_Numbness Awe
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 127/635.
Best 5 of the kept top-20%
R 3.72 · genu 0.15 · blend 0.53 · qual 0.63 · WER 0.00
R 3.65 · genu 0.15 · blend 0.37 · qual 0.72 · WER 0.00
R 3.61 · genu 0.15 · blend 0.69 · qual 0.79 · WER 0.79
R 3.58 · genu 0.20 · blend 0.93 · qual 0.64 · WER 0.76
R 3.56 · genu 0.19 · blend 0.41 · qual 0.59 · WER 0.00
Weakest 5 still inside the top-20% (the selection boundary)
R 1.34 · genu 0.20 · blend 0.35 · qual 0.61 · WER 0.00
R 1.34 · genu 0.17 · blend 1.00 · qual 0.54 · WER 0.84
R 1.33 · genu 0.08 · blend 0.07 · qual 0.72 · WER 0.00
R 1.33 · genu 0.12 · blend 0.15 · qual 0.68 · WER 0.00
R 1.32 · genu 0.17 · blend 0.86 · qual 0.58 · WER 0.70
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.”