A measured merge-scaling guide for the 57 VoiceNet voice-performance dimension LoRAs on
laion/moss-tts-local-transformer-4.55b-voice-acting-v2. For each dimension, both the __high and
__low LoRA are swept at 25 / 50 / 75 / 100 / 125%; every page reports the dose-response of the target
score, the cost in genuineness / vocal-burst blend / speech quality, the strongest ± side-effects on EmoNet emotions
and other VoiceNet dimensions, three audio samples per dose, and a grounded prompting tip. Companion to the
emotion conditioning manual.
Markdown mirror (no audio) for use as an agent skill: md/dims/index.md.