Evolved prompt+setting recipes to push the Sadness LoRA toward the extreme. Fitness combines the EmoNet target (Sadness), vocal-burst blend, genuineness and DSP proxies (loudness, pitch/jitter, spectral flatness for screams, 3–9 Hz amplitude tremor for crying/shivering).
Best fitness 0.740 vs edge-prompt-without-LoRA 0.639 (LoRA lift +0.101) vs neutral 0.241. Winning λ ≈ 0.75; top burst-tags: <voice breaks> (7), <crying> (5), <sobbing> (4), <whimpers> (3).
Method & reproducibility — how the evolution ran
Base model:laion/moss-tts-local-transformer-4.55b-voice-acting-v2 · LoRA: the matching v3 emotion LoRA (rank 32) applied via PEFT adapter scaling at strength λ. Algorithm: genetic search — population 24, 6 generations, elitism (top 8). Each genome is generated best-of-4; its fitness is a multi-objective score for the edge case = the EmoNet target + vocal-burst blend + genuineness + DSP proxies (RMS loudness, median pitch, pitch jitter, spectral flatness for screams, 3–9 Hz amplitude tremor for crying/shivering); the best-fitness candidate represents the genome. Offspring = 60% crossover of two elites + mutation, 40% mutation of one elite. Genome (what evolves): the GENERAL voice description, the SCRIPT cue, the inline vocal-burst tags (e.g. <screams>, <laughs softly>, <sobbing>), the carrier line, the LoRA strength λ, and the sampling params (temperature, top_p, top_k). Mutation operators: 30% new description, 20% add/remove a vocal-burst tag, 15% swap carrier, 17% jitter λ by ±0.25/0.5, then temperature jitter. LoRA on vs off: the two LoRA OFF cards are baselines with the adapter disabled (λ=0) — a neutral prompt and the best edge prompt without the LoRA. Every LoRA ON hall-of-fame recipe was generated with the LoRA active at the evolved λ shown on the card.
GENERAL: A tearful voice, weeping, pushed to the extreme edge of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "No. No, please, not like this."
[sob]<sobbing>
Hall of fame — top 10 evolved recipes (all generated WITH the LoRA active)
GENERAL: A tearful voice, weeping, at the extreme of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "I can't — I can't do this anymore."
GENERAL: A grief-stricken voice, choked with tears, at the extreme of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "I can't — I can't do this anymore."
GENERAL: A broken voice, choked with tears, at the extreme of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "I can't — I can't do this anymore."
GENERAL: A grief-stricken voice, choked with tears, at the extreme of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "I can't — I can't do this anymore."
GENERAL: A mournful voice, choked with tears, at the extreme of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "I can't — I can't do this anymore."
GENERAL: A grief-stricken voice, choked with tears, at the extreme of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "I can't — I can't do this anymore."
GENERAL: A grief-stricken voice, choked with tears, at the extreme of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "I can't — I can't do this anymore."
GENERAL: A broken voice, sobbing, pushed to the extreme edge of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "I can't — I can't do this anymore."
GENERAL: A broken voice, sobbing, pushed to the extreme edge of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "I can't — I can't do this anymore."
GENERAL: A broken voice, weeping, at the extreme of cry.
SCRIPT: (sobbing, voice trembling and breaking through tears) "I can't — I can't do this anymore."
