← all dimensions · markdown (agent skill) · manual home
Focus VoiceNet dimension guide
Social & Interpersonalcode vn_FOCSslot 49
This dimension measures the direction of the speaker’s interpersonal attention. 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.
Focus controls the direction of the speaker’s interpersonal attention. The __high LoRA shifts it +0.35 at 125% (worst cost: burst-blend -0.30); the __low LoRA shifts it -0.22 at 125% (worst cost: speech quality -0.24). The __high direction is the stronger, more reliable knob here.
▲ HIGH — Focus __high
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.365 | — | — | — | — |
| 25% | 0.386 | +0.02 | +0.00 | -0.04 | +0.00 |
| 50% | 0.498 | +0.13 | -0.03 | -0.05 | -0.01 |
| 75% | 0.486 | +0.12 | -0.05 | -0.05 | +0.05 |
| 100% | 0.607 | +0.24 | -0.08 | -0.13 | +0.14 |
| 125% ★ | 0.716 | +0.35 | -0.06 | -0.30 | +0.12 |
EmoNet emotion correlation
Malevolence/Malice+0.49
Elation+0.39
Interest+0.30
Relief-0.62
Contentment-0.44
Fatigue/Exhaustion-0.43
VoiceNet dimension correlation
Storytelling Style+0.89
Dramatic Style+0.80
Authoritative Style+0.78
Disfluency-0.52
Conversational Style-0.49
Vulnerability-0.49
Listen — 3 samples per dose
25%measured target: 0.63 · 0.31 · 0.57
50%measured target: 0.46 · 0.46 · 0.39
75%measured target: 0.62 · 0.67 · 0.40
100%measured target: 0.46 · 0.91 · 0.55
125% ★measured target: 0.70 · 0.64 · 0.67
Merging Focus __high pushes the dimension: at the recommended 125% dose the measured Focus score moves from a 0.36 baseline to 0.72 (a +0.35 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 (-0.30); across the board genuineness -0.06, burst-blend -0.30, speech quality +0.12. Emotionally it reads as Malevolence/Malice (r 0.49), Elation (0.39) and Interest (0.30), and it moves away from Relief, Contentment and Fatigue/Exhaustion. On other voice dimensions it co-moves most with Storytelling Style (0.89), Dramatic Style (0.80) and Authoritative Style (0.78), while pulling down Disfluency, Conversational Style and Vulnerability. Practical call: merge __high at 100–125%; expect Malevolence/Malice and Storytelling Style to rise and burst-blend to fall. If you don’t want Malevolence/Malice bleeding in, offset it with a prompt cue toward Relief to keep the focus shift without the side-effect.
▼ LOW — Focus __low
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.365 | — | — | — | — |
| 25% | 0.302 | -0.06 | +0.00 | -0.02 | -0.01 |
| 50% | 0.317 | -0.05 | -0.02 | -0.03 | +0.07 |
| 75% | 0.157 | -0.21 | +0.00 | -0.17 | -0.29 |
| 100% | 0.247 | -0.12 | +0.00 | -0.18 | -0.12 |
| 125% ★ | 0.149 | -0.22 | +0.01 | -0.21 | -0.24 |
EmoNet emotion correlation
Interest+0.69
Hope/Enthusiasm/Optimism+0.66
Longing+0.57
Emotional Numbness-0.59
Fatigue/Exhaustion-0.46
Pain-0.44
VoiceNet dimension correlation
Storytelling Style+0.80
Emphasis+0.76
Tempo+0.70
Nasal Resonance-0.60
Conversational Style-0.50
Disfluency-0.45
Listen — 3 samples per dose
25%measured target: 0.25 · 0.19 · 0.28
50%measured target: 0.42 · 0.49 · 0.41
75%measured target: 0.14 · 0.22 · 0.13
100%measured target: 0.09 · 0.56 · 0.27
125% ★measured target: 0.12 · 0.05 · 0.18
Merging Focus __low pulls down the dimension: at the recommended 125% dose the measured Focus score moves from a 0.36 baseline to 0.15 (a -0.22 shift). The dose-response is mostly ordered but noisy between doses. That reach is a real cost — the biggest hit is to speech quality (-0.24); across the board genuineness +0.01, burst-blend -0.21, speech quality -0.24. Emotionally it reads as Interest (r 0.69), Hope/Enthusiasm/Optimism (0.66) and Longing (0.57), and it moves away from Emotional Numbness, Fatigue/Exhaustion and Pain. On other voice dimensions it co-moves most with Storytelling Style (0.80), Emphasis (0.76) and Tempo (0.70), while pulling down Nasal Resonance, Conversational Style and Disfluency. Practical call: merge __low at 100–125%; expect Interest and Storytelling Style to rise and speech quality to fall. If you don’t want Interest bleeding in, offset it with a prompt cue toward Emotional Numbness to keep the focus shift without the side-effect.