← all dimensions · markdown (agent skill) · manual home
Valence Shift VoiceNet dimension guide
Temporal Dynamicscode vn_VALSslot 71
This dimension measures the change in emotional polarity across the segment. 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.
Valence Shift controls the change in emotional polarity across the segment. The __high LoRA shifts it +0.30 at 100% (worst cost: burst-blend -0.39); the __low LoRA shifts it -0.07 at 125% (worst cost: speech quality -0.17). The __high direction is the stronger, more reliable knob here.
▲ HIGH — Valence Shift __high
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.173 | — | — | — | — |
| 25% | 0.182 | +0.01 | -0.03 | -0.01 | +0.05 |
| 50% | 0.237 | +0.06 | -0.02 | -0.07 | +0.06 |
| 75% | 0.364 | +0.19 | -0.04 | -0.28 | +0.06 |
| 100% ★ | 0.468 | +0.30 | -0.08 | -0.39 | +0.07 |
| 125% | 0.441 | +0.27 | -0.07 | -0.46 | -0.01 |
EmoNet emotion correlation
Sourness+0.54
Teasing+0.46
Amusement+0.43
Fear-0.47
Longing-0.46
Relief-0.41
VoiceNet dimension correlation
Nasal Resonance+0.81
Mask Resonance+0.78
Valence+0.77
Vocal-burst blend-0.70
Cognitive Load-0.47
Perceived Gender-0.41
Listen — 3 samples per dose
25%measured target: 0.32 · 0.00 · 0.25
50%measured target: 0.18 · 0.21 · 0.36
75%measured target: 0.17 · 0.34 · 0.41
100% ★measured target: 0.62 · 0.51 · 0.34
125%measured target: 0.32 · 0.39 · 0.41
Merging Valence Shift __high drives the dimension: at the recommended 100% dose the measured Valence Shift score moves from a 0.17 baseline to 0.47 (a +0.30 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 (-0.39); across the board genuineness -0.08, burst-blend -0.39, speech quality +0.07. Emotionally it reads as Sourness (r 0.54), Teasing (0.46) and Amusement (0.43), and it moves away from Fear, Longing and Relief. On other voice dimensions it co-moves most with Nasal Resonance (0.81), Mask Resonance (0.78) and Valence (0.77), while pulling down Vocal-burst blend, Cognitive Load and Perceived Gender. Practical call: merge __high at 75–100%; expect Sourness and Nasal Resonance to rise and burst-blend to fall. If you don’t want Sourness bleeding in, offset it with a prompt cue toward Fear to keep the valence shift shift without the side-effect.
▼ LOW — Valence Shift __low
Dose-response
| dose | target | Δtarget | Δgenu | Δblend | Δqual |
|---|
| baseline (0%) | 0.173 | — | — | — | — |
| 25% | 0.137 | -0.04 | +0.03 | -0.03 | -0.01 |
| 50% | 0.115 | -0.06 | +0.00 | -0.02 | -0.02 |
| 75% | 0.168 | -0.01 | +0.03 | -0.03 | -0.08 |
| 100% | 0.209 | +0.04 | +0.05 | -0.06 | -0.17 |
| 125% ★ | 0.105 | -0.07 | +0.03 | -0.16 | -0.17 |
EmoNet emotion correlation
Concentration+0.24
Thankfulness/Gratitude+0.21
Contemplation+0.21
Elation-0.36
Anger-0.28
Impatience and Irritability-0.28
VoiceNet dimension correlation
Warmth+0.42
Whisper-Talk Style+0.38
Esthetics+0.37
Volatility-0.48
Casual Style-0.42
Respiration-0.36
Listen — 3 samples per dose
25%measured target: 0.15 · 0.00 · 0.06
50%measured target: 0.12 · 0.46 · 0.00
75%measured target: 0.12 · 0.00 · 0.04
100%measured target: 0.15 · 0.32 · 0.32
125% ★measured target: 0.20 · 0.08 · 0.00
Merging Valence Shift __low lowers the dimension: at the recommended 125% dose the measured Valence Shift score moves from a 0.17 baseline to 0.10 (a -0.07 shift). The dose-response is mostly ordered but noisy between doses. That reach is a real cost — the biggest hit is to speech quality (-0.17); across the board genuineness +0.03, burst-blend -0.16, speech quality -0.17. Emotionally it reads as Concentration (r 0.24), Thankfulness/Gratitude (0.21) and Contemplation (0.21), and it moves away from Elation, Anger and Impatience and Irritability. On other voice dimensions it co-moves most with Warmth (0.42), Whisper-Talk Style (0.38) and Esthetics (0.37), while pulling down Volatility, Casual Style and Respiration. Practical call: merge __low at 100–125%; expect Concentration and Warmth to rise and speech quality to fall. If you don’t want Concentration bleeding in, offset it with a prompt cue toward Elation to keep the valence shift shift without the side-effect.