← all dimensions · markdown (agent skill) · manual home
Structure VoiceNet dimension guide
Rhythm & Timingcode vn_STRUslot 47
This dimension measures the macro-level organisation of ideas within the clip. 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.
Structure controls the macro-level organisation of ideas within the clip. The __high LoRA shifts it +0.22 at 125% (worst cost: genuineness -0.05); the __low LoRA shifts it -0.18 at 100% (worst cost: speech quality -0.27). The __high direction is the stronger, more reliable knob here.
▲ HIGH — Structure __high
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.271 | — | — | — | — |
| 25% | 0.348 | +0.08 | -0.02 | +0.06 | +0.05 |
| 50% | 0.372 | +0.10 | -0.05 | +0.01 | +0.09 |
| 75% | 0.411 | +0.14 | -0.05 | -0.01 | +0.12 |
| 100% | 0.489 | +0.22 | -0.06 | -0.05 | +0.10 |
| 125% ★ | 0.494 | +0.22 | -0.05 | -0.02 | +0.15 |
EmoNet emotion correlation
Contemplation+0.38
Emotional Numbness+0.30
Concentration+0.28
Fatigue/Exhaustion-0.49
Helplessness-0.48
Relief-0.37
VoiceNet dimension correlation
Smoothness+0.83
Narrator Style+0.80
Esthetics+0.80
Disfluency-0.77
Genuineness-0.74
Casual Style-0.74
Listen — 3 samples per dose
25%measured target: 0.29 · 0.27 · 0.28
50%measured target: 0.38 · 0.55 · 0.52
75%measured target: 0.49 · 0.49 · 0.41
100%measured target: 0.47 · 0.48 · 0.38
125% ★measured target: 0.66 · 0.49 · 0.40
Merging Structure __high pushes the dimension: at the recommended 125% dose the measured Structure score moves from a 0.27 baseline to 0.49 (a +0.22 shift). The dose-response is clean and monotonic — each step pushes the dimension further. That reach is nearly free — the biggest hit is to genuineness (-0.05); across the board genuineness -0.05, burst-blend -0.02, speech quality +0.15. Emotionally it reads as Contemplation (r 0.38), Emotional Numbness (0.30) and Concentration (0.28), and it moves away from Fatigue/Exhaustion, Helplessness and Relief. On other voice dimensions it co-moves most with Smoothness (0.83), Narrator Style (0.80) and Esthetics (0.80), while pulling down Disfluency, Genuineness and Casual Style. Practical call: merge __high at 100–125%; expect Contemplation and Smoothness to rise and genuineness to fall. If you don’t want Contemplation bleeding in, offset it with a prompt cue toward Fatigue/Exhaustion to keep the structure shift without the side-effect.
▼ LOW — Structure __low
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.271 | — | — | — | — |
| 25% | 0.171 | -0.10 | +0.01 | -0.18 | -0.16 |
| 50% | 0.183 | -0.09 | +0.00 | -0.13 | -0.14 |
| 75% | 0.267 | +0.00 | -0.01 | -0.23 | -0.09 |
| 100% ★ | 0.089 | -0.18 | +0.04 | -0.20 | -0.27 |
| 125% | 0.216 | -0.05 | -0.05 | -0.34 | -0.20 |
EmoNet emotion correlation
Interest+0.60
Contentment+0.59
Hope/Enthusiasm/Optimism+0.58
Distress-0.62
Fear-0.58
Pain-0.55
VoiceNet dimension correlation
Esthetics+0.84
Recording Quality+0.84
Smoothness+0.81
Disfluency-0.82
Casual Style-0.56
Genuineness-0.53
Listen — 3 samples per dose
25%measured target: 0.35 · 0.00 · 0.03
50%measured target: 0.00 · 0.35 · 0.27
75%measured target: 0.00 · 0.29 · 0.00
100% ★measured target: 0.06 · 0.04 · 0.13
125%measured target: 0.30 · 0.09 · 0.21
Merging Structure __low suppresses the dimension: at the recommended 100% dose the measured Structure score moves from a 0.27 baseline to 0.09 (a -0.18 shift). The dose-response is noisy and non-monotonic — the merge is an unreliable knob here. That reach is a real cost — the biggest hit is to speech quality (-0.27); across the board genuineness +0.04, burst-blend -0.20, speech quality -0.27. Emotionally it reads as Interest (r 0.60), Contentment (0.59) and Hope/Enthusiasm/Optimism (0.58), and it moves away from Distress, Fear and Pain. On other voice dimensions it co-moves most with Esthetics (0.84), Recording Quality (0.84) and Smoothness (0.81), while pulling down Disfluency, Casual Style and Genuineness. Practical call: merge __low at 75–100%; expect Interest and Esthetics to rise and speech quality to fall. If you don’t want Interest bleeding in, offset it with a prompt cue toward Distress to keep the structure shift without the side-effect.