← all dimensions · markdown (agent skill) · manual home
Cartoonish Style VoiceNet dimension guide
Speaking Stylecode vn_S_CARTslot 91
This dimension measures how cartoonish and exaggerated the voice sounds. Below: the merge-scaling dose-response for the __high and __low LoRA (laion/moss-voicenet-dimension-loras) at 25/50/75/100/125%, the measured target value at each dose, the top-3 ± correlated EmoNet emotions and VoiceNet dimensions, three audio samples per dose, and a grounded prompting tip. Costs are shifts vs. the 24-sample neutral baseline in genuineness, vocal-burst blend and speech quality.
Cartoonish Style controls how cartoonish and exaggerated the voice sounds. The __high LoRA shifts it +0.40 at 125% (worst cost: burst-blend -0.32); the __low LoRA shifts it -0.15 at 50% (worst cost: burst-blend -0.04). The __high direction is the stronger, more reliable knob here.
▲ HIGH — Cartoonish Style __high
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.434 | — | — | — | — |
| 25% | 0.455 | +0.02 | -0.01 | -0.04 | +0.06 |
| 50% | 0.510 | +0.08 | -0.06 | -0.20 | -0.02 |
| 75% | 0.734 | +0.30 | -0.09 | -0.37 | -0.01 |
| 100% | 0.825 | +0.39 | -0.08 | -0.42 | -0.18 |
| 125% ★ | 0.830 | +0.40 | -0.04 | -0.32 | -0.18 |
EmoNet emotion correlation
Triumph+0.59
Elation+0.52
Anger+0.50
Contemplation-0.55
Contentment-0.49
Relief-0.46
VoiceNet dimension correlation
Roughness+0.90
Dramatic Style+0.85
Tension+0.84
Warmth-0.64
Vulnerability-0.63
Teacher/Didactic Style-0.54
Listen — 3 samples per dose
25%measured target: 0.40 · 0.40 · 0.45
50%measured target: 0.51 · 0.77 · 0.25
75%measured target: 0.90 · 0.39 · 0.70
100%measured target: 0.89 · 0.81 · 0.77
125% ★measured target: 0.81 · 0.81 · 0.83
Merging Cartoonish Style __high ramps the dimension: at the recommended 125% dose the measured Cartoonish Style score moves from a 0.43 baseline to 0.83 (a +0.40 shift). The dose-response is clean and monotonic — each step pushes the dimension further. That reach is a steep cost — the biggest hit is to burst-blend (-0.32); across the board genuineness -0.04, burst-blend -0.32, speech quality -0.18. Emotionally it reads as Triumph (r 0.59), Elation (0.52) and Anger (0.50), and it moves away from Contemplation, Contentment and Relief. On other voice dimensions it co-moves most with Roughness (0.90), Dramatic Style (0.85) and Tension (0.84), while pulling down Warmth, Vulnerability and Teacher/Didactic Style. Practical call: merge __high at 100–125%; expect Triumph and Roughness to rise and burst-blend to fall. If you don’t want Triumph bleeding in, offset it with a prompt cue toward Contemplation to keep the cartoonish style shift without the side-effect.
▼ LOW — Cartoonish Style __low
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.434 | — | — | — | — |
| 25% | 0.448 | +0.01 | -0.04 | -0.04 | +0.02 |
| 50% ★ | 0.289 | -0.15 | -0.03 | -0.04 | +0.06 |
| 75% | 0.507 | +0.07 | -0.05 | -0.01 | +0.11 |
| 100% | 0.366 | -0.07 | -0.06 | -0.06 | +0.15 |
| 125% | 0.333 | -0.10 | -0.08 | -0.04 | +0.10 |
EmoNet emotion correlation
Infatuation+0.53
Interest+0.51
Malevolence/Malice+0.50
Concentration-0.31
Embarrassment-0.20
Emotional Numbness-0.20
VoiceNet dimension correlation
Roughness+0.83
Dramatic Style+0.74
Storytelling Style+0.65
Valence-0.57
Mask Resonance-0.56
Conversational Style-0.53
Listen — 3 samples per dose
25%measured target: 0.58 · 0.18 · 0.33
50% ★measured target: 0.31 · 0.20 · 0.14
75%measured target: 0.35 · 0.54 · 0.38
100%measured target: 0.28 · 0.19 · 0.26
125%measured target: 0.20 · 0.21 · 0.45
Merging Cartoonish Style __low suppresses the dimension: at the recommended 50% dose the measured Cartoonish Style score moves from a 0.43 baseline to 0.29 (a -0.15 shift). The dose-response is monotonic but it saturates early, so higher doses add little. That reach is nearly free — the biggest hit is to burst-blend (-0.04); across the board genuineness -0.03, burst-blend -0.04, speech quality +0.06. Emotionally it reads as Infatuation (r 0.53), Interest (0.51) and Malevolence/Malice (0.50), and it moves away from Concentration, Embarrassment and Emotional Numbness. On other voice dimensions it co-moves most with Roughness (0.83), Dramatic Style (0.74) and Storytelling Style (0.65), while pulling down Valence, Mask Resonance and Conversational Style. Practical call: merge __low at 25–75%; expect Infatuation and Roughness to rise and burst-blend to fall. If you don’t want Infatuation bleeding in, offset it with a prompt cue toward Concentration to keep the cartoonish style shift without the side-effect.