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.
elemental
Characteristic dimensions (auto, from its own voice profile): ▲ S_STRY ▲ S_DRAM ▲ EMPH ▲ RANG ▼ S_TECH ▼ S_NEWS ▼ S_WHIS ▼ S_ASMR · emo: Interest Awe Longing
Training set = top-20% by reward = z(genu)+z(blend)+z(quality)−z(WER)+0.7·z(profile-match). Kept 127/636.
Best 5 of the kept top-20%
R 5.39 · genu 0.26 · blend 0.31 · qual 0.62 · WER 0.03
R 5.17 · genu 0.21 · blend 0.36 · qual 0.59 · WER 0.00
R 5.12 · genu 0.26 · blend 0.32 · qual 0.58 · WER 0.00
R 4.53 · genu 0.25 · blend 0.38 · qual 0.50 · WER 0.00
R 4.39 · genu 0.24 · blend 0.63 · qual 0.34 · WER 0.00
Weakest 5 still inside the top-20% (the selection boundary)
R 1.33 · genu 0.22 · blend 0.02 · qual 0.36 · WER 0.00
R 1.32 · genu 0.14 · blend 0.04 · qual 0.66 · WER 0.04
R 1.30 · genu 0.21 · blend 0.09 · qual 0.58 · WER 0.00
R 1.28 · genu 0.15 · blend 0.28 · qual 0.26 · WER 0.03
R 1.27 · genu 0.15 · blend 0.10 · qual 0.54 · 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.”