← all dimensions · markdown (agent skill) · manual home
Esthetics VoiceNet dimension guide
Spectral & Timbralcode vn_ESTHslot 67
This dimension measures the overall attractiveness or pleasantness of the sound. 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.
Esthetics controls the overall attractiveness or pleasantness of the sound. The __high LoRA shifts it +0.10 at 125% (worst cost: burst-blend -0.09); the __low LoRA shifts it -0.25 at 100% (worst cost: burst-blend -0.31). The __low direction is the stronger, more reliable knob here.
▲ HIGH — Esthetics __high
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.464 | — | — | — | — |
| 25% | 0.529 | +0.07 | -0.03 | -0.06 | +0.06 |
| 50% | 0.478 | +0.01 | -0.02 | -0.08 | +0.02 |
| 75% | 0.520 | +0.06 | -0.05 | -0.03 | +0.08 |
| 100% | 0.477 | +0.01 | -0.07 | +0.01 | +0.05 |
| 125% ★ | 0.567 | +0.10 | -0.08 | -0.09 | +0.09 |
EmoNet emotion correlation
Contemplation+0.52
Emotional Numbness+0.38
Concentration+0.28
Confusion-0.45
Doubt-0.45
Impatience and Irritability-0.43
VoiceNet dimension correlation
Whisper-Talk Style+0.82
Recording Quality+0.78
Smoothness+0.74
Ranting/Angry Style-0.61
Volatility-0.45
Disfluency-0.44
Listen — 3 samples per dose
25%measured target: 0.52 · 0.48 · 0.51
50%measured target: 0.59 · 0.18 · 0.54
75%measured target: 0.54 · 0.75 · 0.61
100%measured target: 0.58 · 0.63 · 0.56
125% ★measured target: 0.58 · 0.55 · 0.71
Merging Esthetics __high pushes the dimension: at the recommended 125% dose the measured Esthetics score moves from a 0.46 baseline to 0.57 (a +0.10 shift). The dose-response is mostly ordered but noisy between doses. That reach is a modest cost — the biggest hit is to burst-blend (-0.09); across the board genuineness -0.08, burst-blend -0.09, speech quality +0.09. Emotionally it reads as Contemplation (r 0.52), Emotional Numbness (0.38) and Concentration (0.28), and it moves away from Confusion, Doubt and Impatience and Irritability. On other voice dimensions it co-moves most with Whisper-Talk Style (0.82), Recording Quality (0.78) and Smoothness (0.74), while pulling down Ranting/Angry Style, Volatility and Disfluency. Practical call: merge __high at 100–125%; expect Contemplation and Whisper-Talk Style to rise and burst-blend to fall. If you don’t want Contemplation bleeding in, offset it with a prompt cue toward Confusion to keep the esthetics shift without the side-effect.
▼ LOW — Esthetics __low
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.464 | — | — | — | — |
| 25% | 0.492 | +0.03 | -0.03 | -0.12 | +0.02 |
| 50% | 0.413 | -0.05 | -0.01 | -0.25 | -0.05 |
| 75% | 0.234 | -0.23 | +0.03 | -0.22 | -0.21 |
| 100% ★ | 0.217 | -0.25 | +0.01 | -0.31 | -0.22 |
| 125% | 0.248 | -0.22 | -0.02 | -0.32 | -0.22 |
EmoNet emotion correlation
Contemplation+0.77
Interest+0.76
Affection+0.75
Pain-0.75
Distress-0.71
Fear-0.68
VoiceNet dimension correlation
Smoothness+0.94
Recording Quality+0.93
Whisper-Talk Style+0.91
Disfluency-0.76
Playful Style-0.75
Dynamic Arc-0.73
Listen — 3 samples per dose
25%measured target: 0.33 · 0.56 · 0.68
50%measured target: 0.06 · 0.44 · 0.73
75%measured target: 0.53 · 0.09 · 0.44
100% ★measured target: 0.00 · 0.09 · 0.65
125%measured target: 0.45 · 0.05 · 0.14
Merging Esthetics __low pulls down the dimension: at the recommended 100% dose the measured Esthetics score moves from a 0.46 baseline to 0.22 (a -0.25 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 (-0.31); across the board genuineness +0.01, burst-blend -0.31, speech quality -0.22. Emotionally it reads as Contemplation (r 0.77), Interest (0.76) and Affection (0.75), and it moves away from Pain, Distress and Fear. On other voice dimensions it co-moves most with Smoothness (0.94), Recording Quality (0.93) and Whisper-Talk Style (0.91), while pulling down Disfluency, Playful Style and Dynamic Arc. Practical call: merge __low at 75–100%; expect Contemplation and Smoothness to rise and burst-blend to fall. If you don’t want Contemplation bleeding in, offset it with a prompt cue toward Pain to keep the esthetics shift without the side-effect.