← all dimensions · markdown (agent skill) · manual home
Background Noise VoiceNet dimension guide
Language & Recordingcode vn_BKGNslot 73
This dimension measures the prominence and intrusiveness of ambient 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.
Background Noise controls the prominence and intrusiveness of ambient sound. The __high LoRA shifts it +0.08 at 100% (worst cost: genuineness -0.04); the __low LoRA shifts it -0.09 at 50% (worst cost: speech quality -0.15). The __low direction is the stronger, more reliable knob here.
▲ HIGH — Background Noise __high
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.399 | — | — | — | — |
| 25% | 0.380 | -0.02 | -0.01 | -0.01 | +0.01 |
| 50% | 0.381 | -0.02 | -0.02 | +0.07 | -0.09 |
| 75% | 0.446 | +0.05 | -0.02 | -0.01 | +0.02 |
| 100% ★ | 0.478 | +0.08 | -0.04 | +0.02 | +0.09 |
| 125% | 0.432 | +0.03 | +0.00 | +0.03 | -0.08 |
EmoNet emotion correlation
Hope/Enthusiasm/Optimism+0.47
Contentment+0.40
Interest+0.35
Fatigue/Exhaustion-0.65
Distress-0.45
Helplessness-0.42
VoiceNet dimension correlation
Mixed Resonance+0.68
Register+0.67
Mask Resonance+0.64
Disfluency-0.63
Perceived Gender-0.57
Cognitive Load-0.52
Listen — 3 samples per dose
25%measured target: 0.41 · 0.40 · 0.34
50%measured target: 0.43 · 0.26 · 0.32
75%measured target: 0.37 · 0.55 · 0.55
100% ★measured target: 0.50 · 0.40 · 0.52
125%measured target: 0.40 · 0.54 · 0.44
Merging Background Noise __high drives the dimension: at the recommended 100% dose the measured Background Noise score moves from a 0.40 baseline to 0.48 (a +0.08 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 (-0.04); across the board genuineness -0.04, burst-blend +0.02, speech quality +0.09. Emotionally it reads as Hope/Enthusiasm/Optimism (r 0.47), Contentment (0.40) and Interest (0.35), and it moves away from Fatigue/Exhaustion, Distress and Helplessness. On other voice dimensions it co-moves most with Mixed Resonance (0.68), Register (0.67) and Mask Resonance (0.64), while pulling down Disfluency, Perceived Gender and Cognitive Load. Practical call: merge __high at 75–100%; expect Hope/Enthusiasm/Optimism and Mixed Resonance to rise and genuineness to fall. If you don’t want Hope/Enthusiasm/Optimism bleeding in, offset it with a prompt cue toward Fatigue/Exhaustion to keep the background noise shift without the side-effect.
▼ LOW — Background Noise __low
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.399 | — | — | — | — |
| 25% | 0.440 | +0.04 | -0.02 | -0.06 | +0.04 |
| 50% ★ | 0.305 | -0.09 | +0.01 | -0.10 | -0.15 |
| 75% | 0.389 | -0.01 | +0.01 | -0.15 | -0.11 |
| 100% | 0.374 | -0.02 | -0.02 | -0.28 | -0.10 |
| 125% | 0.313 | -0.09 | +0.01 | -0.38 | -0.20 |
EmoNet emotion correlation
Contentment+0.49
Hope/Enthusiasm/Optimism+0.44
Interest+0.44
Fatigue/Exhaustion-0.52
Distress-0.42
Pain-0.42
VoiceNet dimension correlation
Recording Quality+0.71
Register+0.64
Narrator Style+0.60
Disfluency-0.64
Perceived Gender-0.50
Genuineness-0.42
Listen — 3 samples per dose
25%measured target: 0.40 · 0.56 · 0.41
50% ★measured target: 0.29 · 0.27 · 0.30
75%measured target: 0.38 · 0.22 · 0.38
100%measured target: 0.36 · 0.30 · 0.36
125%measured target: 0.39 · 0.27 · 0.28
Merging Background Noise __low lowers the dimension: at the recommended 50% dose the measured Background Noise score moves from a 0.40 baseline to 0.30 (a -0.09 shift). The dose-response is monotonic but it saturates early, so higher doses add little. That reach is a real cost — the biggest hit is to speech quality (-0.15); across the board genuineness +0.01, burst-blend -0.10, speech quality -0.15. Emotionally it reads as Contentment (r 0.49), Hope/Enthusiasm/Optimism (0.44) and Interest (0.44), and it moves away from Fatigue/Exhaustion, Distress and Pain. On other voice dimensions it co-moves most with Recording Quality (0.71), Register (0.64) and Narrator Style (0.60), while pulling down Disfluency, Perceived Gender and Genuineness. Practical call: merge __low at 25–75%; expect Contentment and Recording Quality to rise and speech quality to fall. If you don’t want Contentment bleeding in, offset it with a prompt cue toward Fatigue/Exhaustion to keep the background noise shift without the side-effect.