← all dimensions · markdown (agent skill) · manual home
Formal Style VoiceNet dimension guide
Speaking Stylecode vn_S_FORMslot 84
This dimension measures how formal and composed the delivery 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.
Formal Style controls how formal and composed the delivery sounds. The __high LoRA shifts it +0.44 at 125% (worst cost: burst-blend -0.23); the __low LoRA shifts it -0.10 at 100% (worst cost: burst-blend -0.35). The __high direction is the stronger, more reliable knob here.
▲ HIGH — Formal Style __high
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.261 | — | — | — | — |
| 25% | 0.370 | +0.11 | -0.05 | -0.05 | +0.10 |
| 50% | 0.486 | +0.22 | -0.08 | -0.08 | +0.11 |
| 75% | 0.522 | +0.26 | -0.08 | -0.09 | +0.14 |
| 100% | 0.654 | +0.39 | -0.12 | -0.13 | +0.20 |
| 125% ★ | 0.696 | +0.44 | -0.11 | -0.23 | +0.19 |
EmoNet emotion correlation
Emotional Numbness+0.50
Sexual Lust+0.33
Concentration+0.32
Relief-0.69
Helplessness-0.41
Fatigue/Exhaustion-0.40
VoiceNet dimension correlation
Structure+0.90
Smoothness+0.87
Narrator Style+0.86
Casual Style-0.91
Genuineness-0.78
Disfluency-0.74
Listen — 3 samples per dose
25%measured target: 0.17 · 0.57 · 0.23
50%measured target: 0.58 · 0.54 · 0.35
75%measured target: 0.57 · 0.64 · 0.39
100%measured target: 0.52 · 0.65 · 0.71
125% ★measured target: 0.75 · 0.58 · 0.68
Merging Formal Style __high pushes the dimension: at the recommended 125% dose the measured Formal Style score moves from a 0.26 baseline to 0.70 (a +0.44 shift). The dose-response is clean and monotonic — each step pushes the dimension further. That reach is a real cost — the biggest hit is to burst-blend (-0.23); across the board genuineness -0.11, burst-blend -0.23, speech quality +0.19. Emotionally it reads as Emotional Numbness (r 0.50), Sexual Lust (0.33) and Concentration (0.32), and it moves away from Relief, Helplessness and Fatigue/Exhaustion. On other voice dimensions it co-moves most with Structure (0.90), Smoothness (0.87) and Narrator Style (0.86), while pulling down Casual Style, Genuineness and Disfluency. Practical call: merge __high at 100–125%; expect Emotional Numbness and Structure to rise and burst-blend to fall. If you don’t want Emotional Numbness bleeding in, offset it with a prompt cue toward Relief to keep the formal style shift without the side-effect.
▼ LOW — Formal Style __low
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.261 | — | — | — | — |
| 25% | 0.280 | +0.02 | -0.03 | -0.09 | -0.02 |
| 50% | 0.380 | +0.12 | -0.05 | -0.07 | +0.03 |
| 75% | 0.189 | -0.07 | +0.03 | -0.14 | -0.19 |
| 100% ★ | 0.165 | -0.10 | -0.06 | -0.35 | -0.18 |
| 125% | 0.205 | -0.06 | -0.09 | -0.31 | -0.14 |
EmoNet emotion correlation
Longing+0.45
Interest+0.39
Infatuation+0.39
Pain-0.47
Impatience and Irritability-0.40
Astonishment/Surprise-0.39
VoiceNet dimension correlation
Narrator Style+0.76
Smoothness+0.74
Fullness+0.71
Casual Style-0.77
Respiration-0.65
Volatility-0.60
Listen — 3 samples per dose
25%measured target: 0.09 · 0.34 · 0.15
50%measured target: 0.20 · 0.43 · 0.67
75%measured target: 0.03 · 0.30 · 0.02
100% ★measured target: 0.01 · 0.09 · 0.36
125%measured target: 0.27 · 0.29 · 0.03
Merging Formal Style __low lowers the dimension: at the recommended 100% dose the measured Formal Style score moves from a 0.26 baseline to 0.17 (a -0.10 shift). The dose-response is mostly ordered but noisy between doses. That reach is a steep cost — the biggest hit is to burst-blend (-0.35); across the board genuineness -0.06, burst-blend -0.35, speech quality -0.18. Emotionally it reads as Longing (r 0.45), Interest (0.39) and Infatuation (0.39), and it moves away from Pain, Impatience and Irritability and Astonishment/Surprise. On other voice dimensions it co-moves most with Narrator Style (0.76), Smoothness (0.74) and Fullness (0.71), while pulling down Casual Style, Respiration and Volatility. Practical call: merge __low at 75–100%; expect Longing and Narrator Style to rise and burst-blend to fall. If you don’t want Longing bleeding in, offset it with a prompt cue toward Pain to keep the formal style shift without the side-effect.