← all dimensions · markdown (agent skill) · manual home
Cognitive Load VoiceNet dimension guide
Physical Productioncode vn_COGLslot 59
This dimension measures the mental friction involved in formulating speech. 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.
Cognitive Load controls the mental friction involved in formulating speech. The __high LoRA shifts it +0.08 at 100% (worst cost: speech quality -0.11); the __low LoRA shifts it -0.17 at 100% (worst cost: burst-blend -0.33). The __low direction is the stronger, more reliable knob here.
▲ HIGH — Cognitive Load __high
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.322 | — | — | — | — |
| 25% | 0.309 | -0.01 | -0.02 | -0.06 | -0.03 |
| 50% | 0.323 | +0.00 | -0.02 | -0.01 | +0.00 |
| 75% | 0.333 | +0.01 | +0.02 | +0.07 | -0.07 |
| 100% ★ | 0.398 | +0.08 | +0.07 | +0.01 | -0.11 |
| 125% | 0.332 | +0.01 | +0.03 | +0.09 | -0.05 |
EmoNet emotion correlation
Sadness+0.53
Fear+0.49
Helplessness+0.47
Contentment-0.52
Pleasure/Ecstasy-0.42
Hope/Enthusiasm/Optimism-0.37
VoiceNet dimension correlation
Disfluency+0.72
Tension+0.64
Genuineness+0.57
Recording Quality-0.72
Teacher/Didactic Style-0.69
Velocity Flux-0.66
Listen — 3 samples per dose
25%measured target: 0.25 · 0.47 · 0.49
50%measured target: 0.33 · 0.28 · 0.46
75%measured target: 0.45 · 0.34 · 0.23
100% ★measured target: 0.30 · 0.46 · 0.41
125%measured target: 0.37 · 0.26 · 0.47
Merging Cognitive Load __high lifts the dimension: at the recommended 100% dose the measured Cognitive Load score moves from a 0.32 baseline to 0.40 (a +0.08 shift). The dose-response is clean and monotonic — each step pushes the dimension further. That reach is a modest cost — the biggest hit is to speech quality (-0.11); across the board genuineness +0.07, burst-blend +0.01, speech quality -0.11. Emotionally it reads as Sadness (r 0.53), Fear (0.49) and Helplessness (0.47), and it moves away from Contentment, Pleasure/Ecstasy and Hope/Enthusiasm/Optimism. On other voice dimensions it co-moves most with Disfluency (0.72), Tension (0.64) and Genuineness (0.57), while pulling down Recording Quality, Teacher/Didactic Style and Velocity Flux. Practical call: merge __high at 75–100%; expect Sadness and Disfluency to rise and speech quality to fall. If you don’t want Sadness bleeding in, offset it with a prompt cue toward Contentment to keep the cognitive load shift without the side-effect.
▼ LOW — Cognitive Load __low
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.322 | — | — | — | — |
| 25% | 0.262 | -0.06 | -0.04 | -0.08 | -0.06 |
| 50% | 0.280 | -0.04 | -0.07 | -0.05 | +0.05 |
| 75% | 0.171 | -0.15 | -0.04 | -0.08 | -0.02 |
| 100% ★ | 0.150 | -0.17 | -0.10 | -0.33 | +0.09 |
| 125% | 0.181 | -0.14 | -0.10 | -0.37 | -0.01 |
EmoNet emotion correlation
Fatigue/Exhaustion+0.38
Longing+0.33
Helplessness+0.29
Contempt-0.41
Sourness-0.40
Teasing-0.30
VoiceNet dimension correlation
Disfluency+0.70
Vulnerability+0.55
Perceived Gender+0.47
Brightness-0.69
Background Noise-0.67
Chunking-0.66
Listen — 3 samples per dose
25%measured target: 0.22 · 0.23 · 0.38
50%measured target: 0.24 · 0.53 · 0.33
75%measured target: 0.19 · 0.27 · 0.04
100% ★measured target: 0.15 · 0.11 · 0.06
125%measured target: 0.08 · 0.20 · 0.11
Merging Cognitive Load __low suppresses the dimension: at the recommended 100% dose the measured Cognitive Load score moves from a 0.32 baseline to 0.15 (a -0.17 shift). The dose-response is mostly ordered but noisy between doses. That reach is a steep cost — the biggest hit is to burst-blend (-0.33); across the board genuineness -0.10, burst-blend -0.33, speech quality +0.09. Emotionally it reads as Fatigue/Exhaustion (r 0.38), Longing (0.33) and Helplessness (0.29), and it moves away from Contempt, Sourness and Teasing. On other voice dimensions it co-moves most with Disfluency (0.70), Vulnerability (0.55) and Perceived Gender (0.47), while pulling down Brightness, Background Noise and Chunking. Practical call: merge __low at 75–100%; expect Fatigue/Exhaustion and Disfluency to rise and burst-blend to fall. If you don’t want Fatigue/Exhaustion bleeding in, offset it with a prompt cue toward Contempt to keep the cognitive load shift without the side-effect.