The merged parent used in these specialist comparisons is now released as Humaneness Voice Large 1.6 Alpha: 1.5 + final RealSpeech rank 4, merged once. Apply only the selected new bucket adapter. Do not reload either ancestor.
1.5 Original model · 1.6 merged model · Complete context manuals for humans and agents
Read 1.5 context manual · Read 1.6 + adapter context manual. Exact prompts, CFG, adapter scales and failed gates are retained. New model/documentation contributions: CC BY 4.0; existing upstream and gated specialist terms remain applicable.
The original grid used generic GENERAL/SCRIPT directions at CFG1. It did not systematically use target-specific manual winners. The new study freezes the actual rank-4 manual wording, preserves every literal word, seed, reference, adapter and frame budget, then independently varies CFG and adapter strength. The previous recordings and conclusions remain below.
Ranking:100% requested EARS emotion/style head, standardized using frozen training statistics; low-pole targets reverse the sign. Formal/casual contrasts average the desired head and negative opposite head. CE and genuineness are diagnostics. Higher target z means stronger desired delivery, not a human rating. More CFG is not automatically better.
Selection:20 original-text outcomes per configuration select wording, CFG and scale. Fifty additional unseen-text reference pairs test the frozen choices. A recommended adapter must improve against both the same-CFG no-adapter control and the best manual-only setting. Quality gates: per-take WER≤15%, mean WER≤10%, valid/terminated fraction≥90%, paired identity change≥−0.05 over50pairs. Failed outcomes stay visible. Language-cluster bootstrap intervals use only five fixed reference voices; findings across72targets remain exploratory.
Five languages: EN/DE/FR/NL/ES; original seeds1701/1702; reference187frames/14.96seconds. CFG1/1.25/1.5/2/2.5/3/4/5 × bucket strength0/0.5/0.75/1. CFG4/5 are stress conditions, not presumed improvements. No-reference use of these bucket recommendations has not been validated.
Taxonomy:S_MONO means Monologue, S_TECH means Teacher/Didactic, VOLT means Volatility. The two corrected style recipes and17dimension/contrast recipes are new searches, not historical manual winners.53other targets reuse directly applicable manual recipes.
Best tested agent route: No new configuration passes the held-out gates; retain standard fallback and inspect audio. Bucket scale 0.0, CFG 1.0.
| Held-out condition | n | Target z↑ | Raw head | CE↑ | Genuine↑ | Voice↑ | WER↓ | Valid | QA |
|---|---|---|---|---|---|---|---|---|---|
| standard | 50 | -0.322 | 0.064 | 6.241 | 4.390 | 0.571 | 0.009 | 1.000 | pass |
| manual | 50 | -0.292 | 0.077 | 6.210 | 4.102 | 0.502 | 0.009 | 1.000 | fail |
| adapter | 50 | -0.310 | 0.069 | 6.253 | 4.216 | 0.558 | 0.005 | 1.000 | pass |
| matched_no_adapter | 50 | -0.337 | 0.057 | 6.270 | 4.220 | 0.561 | 0.007 | 1.000 | pass |
Incremental adapter gain versus same-CFG control: 0.027 z;95% cluster interval [0.0012890543250093658, 0.07433593510530578]. Versus best manual-only: -0.018 z. Manual-only versus standard: 0.030 z.
| Scale | CFG | Target z | CE | Genuine | WER | Provisional QA |
|---|---|---|---|---|---|---|
| 0.0 | 1.0 | -0.393 | 6.200 | 4.239 | 0.003 | False |
| 0.5 | 1.0 | -0.401 | 6.206 | 4.182 | 0.017 | True |
| 0.75 | 1.0 | -0.388 | 6.244 | 4.295 | 0.000 | True |
| 1.0 | 1.0 | -0.403 | 6.219 | 4.226 | 0.003 | True |
| 0.0 | 1.25 | -0.401 | 6.158 | 4.201 | 0.005 | True |
| 0.5 | 1.25 | -0.389 | 6.184 | 4.178 | 0.006 | False |
| 0.75 | 1.25 | -0.404 | 6.221 | 4.206 | 0.000 | True |
| 1.0 | 1.25 | -0.389 | 6.195 | 4.188 | 0.003 | False |
| 0.0 | 1.5 | -0.398 | 6.202 | 4.133 | 0.000 | True |
| 0.5 | 1.5 | -0.166 | 6.183 | 4.194 | 0.000 | True |
| 0.75 | 1.5 | -0.381 | 6.201 | 4.208 | 0.003 | True |
| 1.0 | 1.5 | -0.250 | 6.200 | 4.200 | 0.000 | False |
| 0.0 | 2.0 | -0.291 | 6.154 | 4.134 | 0.000 | False |
| 0.5 | 2.0 | -0.360 | 6.196 | 4.177 | 0.011 | False |
| 0.75 | 2.0 | -0.236 | 6.185 | 4.114 | 0.000 | False |
| 1.0 | 2.0 | -0.169 | 6.143 | 4.092 | 0.003 | False |
| 0.0 | 2.5 | -0.248 | 6.146 | 4.145 | 0.000 | False |
| 0.5 | 2.5 | -0.212 | 6.152 | 4.051 | 0.003 | False |
| 0.75 | 2.5 | -0.151 | 6.098 | 4.088 | 0.000 | False |
| 1.0 | 2.5 | -0.263 | 6.163 | 4.098 | 0.000 | False |
| 0.0 | 3.0 | -0.158 | 6.082 | 4.061 | 0.021 | True |
| 0.5 | 3.0 | -0.034 | 5.965 | 4.026 | 0.050 | False |
| 0.75 | 3.0 | -0.176 | 6.024 | 4.083 | 0.006 | False |
| 1.0 | 3.0 | -0.063 | 6.008 | 4.106 | 0.024 | False |
| 0.0 | 4.0 | -0.142 | 6.070 | 4.100 | 0.006 | False |
| 0.5 | 4.0 | -0.240 | 5.984 | 3.936 | 0.027 | False |
| 0.75 | 4.0 | -0.135 | 6.022 | 4.001 | 0.006 | False |
| 1.0 | 4.0 | 0.210 | 6.056 | 3.987 | 0.013 | False |
| 0.0 | 5.0 | 0.313 | 5.986 | 4.010 | 0.003 | False |
| 0.5 | 5.0 | 0.189 | 5.971 | 3.950 | 0.003 | False |
| 0.75 | 5.0 | 0.544 | 6.044 | 3.977 | 0.009 | False |
| 1.0 | 5.0 | 0.112 | 6.012 | 3.966 | 0.015 | False |
direct_manual_recipe
{
"id": "manual-0f7cf1dc94a9",
"general": "Focus on precise articulation, exaggerating the crispness of consonants to convey intellectual superiority and disdain.",
"inline": [
"Over-articulate the plosives in the initial phrase with sharp, percussive stops.",
"Allow vowel sounds to be slightly clipped, as if speaking through a filter of utter dismissal.",
"Ensure the final consonants have a distinct, brittle quality that cuts through the air."
],
"cfg": 2,
"recipe": "P08",
"strategy": "local-Gemma-mutation",
"timing_syntax": "canonical",
"canonical_labels": true,
"neutral_event": "keep",
"parent": "cfg-2",
"origin": "exact_frozen_manual_recipe"
}Exact machine-readable comparison · Agent tool recommendations
Intense Contempt. Base: 1.5 Alpha + merged real-speech rank4; new adapter: rank32. Every row is a matched seed/text/reference test. Scale zero disables the new adapter; all other scales are applied absolutely, never compounded.
Adapter training data: {"de": 22, "fr": 20, "en": 50, "es": 4, "nl": 1}. All available accepted languages were pooled, globally shuffled, for three epochs. Peak LR 2e-05 at the end of the first-epoch warmup, then cosine decay. All original optimizer/RNG checkpoints remain saved.
Taxonomy: S_MONO is monologue, not monotone; VOLT is volatility, not volume. No claim is made that a classifier score alone proves the audible performance.
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 20 | 0.353 ± 0.045 | 6.170 ± 0.148 | 4.359 ± 0.243 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 20 | 0.356 ± 0.043 | 6.178 ± 0.128 | 4.337 ± 0.229 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 20 | 0.348 ± 0.043 | 6.164 ± 0.125 | 4.361 ± 0.217 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 20 | 0.354 ± 0.050 | 6.197 ± 0.152 | 4.269 ± 0.264 | 0.00% ± 0.00 |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.322 ± 0.068 | 6.239 ± 0.139 | 4.264 ± 0.247 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.331 ± 0.062 | 6.277 ± 0.084 | 4.285 ± 0.331 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.315 ± 0.055 | 6.293 ± 0.071 | 4.211 ± 0.268 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.319 ± 0.072 | 6.283 ± 0.101 | 4.168 ± 0.274 | 0.00% ± 0.00 |
| Prompt / seed | Merged base · no new LoRA | Rank32 × 0.5 | Rank32 × 0.75 | Rank32 × 1.0 |
|---|---|---|---|---|
| Seed 1701 The door is still open. I put your coat on the chair. Tell me when you are ready. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] The door is still open. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] I put your coat on the chair. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Tell me when you are ready. | Target 0.377 · CE 6.22 · Genuine 4.03 · WER 0.0% | Target 0.380 · CE 6.17 · Genuine 4.01 · WER 0.0% | Target 0.341 · CE 6.29 · Genuine 4.13 · WER 0.0% | Target 0.375 · CE 6.17 · Genuine 3.93 · WER 0.0% |
| Seed 1702 The door is still open. I put your coat on the chair. Tell me when you are ready. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] The door is still open. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] I put your coat on the chair. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Tell me when you are ready. | Target 0.384 · CE 6.05 · Genuine 4.09 · WER 0.0% | Target 0.386 · CE 6.28 · Genuine 4.01 · WER 0.0% | Target 0.380 · CE 6.21 · Genuine 3.87 · WER 0.0% | Target 0.387 · CE 6.22 · Genuine 3.96 · WER 0.0% |
| Seed 1701 I found the little box. It was beside the old window. We can look inside together. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] I found the little box. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] It was beside the old window. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] We can look inside together. | Target 0.248 · CE 6.38 · Genuine 4.56 · WER 0.0% | Target 0.258 · CE 6.38 · Genuine 4.68 · WER 0.0% | Target 0.272 · CE 6.38 · Genuine 4.40 · WER 0.0% | Target 0.258 · CE 6.37 · Genuine 4.51 · WER 0.0% |
| Seed 1702 I found the little box. It was beside the old window. We can look inside together. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] I found the little box. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] It was beside the old window. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] We can look inside together. | Target 0.280 · CE 6.30 · Genuine 4.38 · WER 0.0% | Target 0.300 · CE 6.28 · Genuine 4.43 · WER 0.0% | Target 0.266 · CE 6.29 · Genuine 4.44 · WER 0.0% | Target 0.256 · CE 6.36 · Genuine 4.28 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.389 ± 0.017 | 6.322 ± 0.091 | 4.529 ± 0.074 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.385 ± 0.017 | 6.257 ± 0.104 | 4.476 ± 0.198 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.386 ± 0.010 | 6.261 ± 0.093 | 4.452 ± 0.122 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.393 ± 0.017 | 6.344 ± 0.106 | 4.430 ± 0.113 | 0.00% ± 0.00 |
| Prompt / seed | Merged base · no new LoRA | Rank32 × 0.5 | Rank32 × 0.75 | Rank32 × 1.0 |
|---|---|---|---|---|
| Seed 1701 Die Tür ist noch offen. Ich habe deinen Mantel auf den Stuhl gelegt. Sag mir, wenn du bereit bist. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Die Tür ist noch offen. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Ich habe deinen Mantel auf den Stuhl gelegt. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Sag mir, wenn du bereit bist. | Target 0.382 · CE 6.19 · Genuine 4.56 · WER 0.0% | Target 0.393 · CE 6.11 · Genuine 4.67 · WER 0.0% | Target 0.388 · CE 6.14 · Genuine 4.39 · WER 0.0% | Target 0.407 · CE 6.24 · Genuine 4.48 · WER 0.0% |
| Seed 1702 Die Tür ist noch offen. Ich habe deinen Mantel auf den Stuhl gelegt. Sag mir, wenn du bereit bist. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Die Tür ist noch offen. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Ich habe deinen Mantel auf den Stuhl gelegt. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Sag mir, wenn du bereit bist. | Target 0.400 · CE 6.34 · Genuine 4.62 · WER 0.0% | Target 0.370 · CE 6.28 · Genuine 4.47 · WER 0.0% | Target 0.398 · CE 6.29 · Genuine 4.63 · WER 0.0% | Target 0.377 · CE 6.28 · Genuine 4.51 · WER 0.0% |
| Seed 1701 Ich habe die kleine Schachtel gefunden. Sie stand neben dem alten Fenster. Wir können gemeinsam hineinschauen. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Ich habe die kleine Schachtel gefunden. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Sie stand neben dem alten Fenster. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Wir können gemeinsam hineinschauen. | Target 0.368 · CE 6.39 · Genuine 4.48 · WER 0.0% | Target 0.405 · CE 6.36 · Genuine 4.21 · WER 0.0% | Target 0.377 · CE 6.25 · Genuine 4.42 · WER 0.0% | Target 0.408 · CE 6.48 · Genuine 4.47 · WER 0.0% |
| Seed 1702 Ich habe die kleine Schachtel gefunden. Sie stand neben dem alten Fenster. Wir können gemeinsam hineinschauen. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Ich habe die kleine Schachtel gefunden. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Sie stand neben dem alten Fenster. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Wir können gemeinsam hineinschauen. | Target 0.405 · CE 6.37 · Genuine 4.46 · WER 0.0% | Target 0.372 · CE 6.28 · Genuine 4.56 · WER 0.0% | Target 0.380 · CE 6.36 · Genuine 4.36 · WER 0.0% | Target 0.382 · CE 6.38 · Genuine 4.26 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.338 ± 0.035 | 6.174 ± 0.034 | 4.514 ± 0.186 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.321 ± 0.029 | 6.150 ± 0.098 | 4.302 ± 0.246 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.337 ± 0.046 | 6.086 ± 0.073 | 4.423 ± 0.180 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.322 ± 0.021 | 6.118 ± 0.143 | 4.258 ± 0.280 | 0.00% ± 0.00 |
| Prompt / seed | Merged base · no new LoRA | Rank32 × 0.5 | Rank32 × 0.75 | Rank32 × 1.0 |
|---|---|---|---|---|
| Seed 1701 La porte est encore ouverte. J’ai posé ton manteau sur la chaise. Dis-moi quand tu es prêt. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] La porte est encore ouverte. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] J’ai posé ton manteau sur la chaise. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Dis-moi quand tu es prêt. | Target 0.386 · CE 6.17 · Genuine 4.42 · WER 0.0% | Target 0.333 · CE 6.01 · Genuine 3.94 · WER 0.0% | Target 0.353 · CE 6.01 · Genuine 4.41 · WER 0.0% | Target 0.333 · CE 5.95 · Genuine 3.87 · WER 0.0% |
| Seed 1702 La porte est encore ouverte. J’ai posé ton manteau sur la chaise. Dis-moi quand tu es prêt. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] La porte est encore ouverte. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] J’ai posé ton manteau sur la chaise. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Dis-moi quand tu es prêt. | Target 0.340 · CE 6.13 · Genuine 4.34 · WER 0.0% | Target 0.357 · CE 6.19 · Genuine 4.35 · WER 0.0% | Target 0.384 · CE 6.04 · Genuine 4.22 · WER 0.0% | Target 0.342 · CE 6.05 · Genuine 4.25 · WER 0.0% |
| Seed 1701 J’ai trouvé la petite boîte. Elle était près de la vieille fenêtre. Nous pouvons regarder ensemble ce qu’elle contient. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] J’ai trouvé la petite boîte. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Elle était près de la vieille fenêtre. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Nous pouvons regarder ensemble ce qu’elle contient. | Target 0.313 · CE 6.21 · Genuine 4.52 · WER 0.0% | Target 0.296 · CE 6.17 · Genuine 4.48 · WER 0.0% | Target 0.274 · CE 6.12 · Genuine 4.41 · WER 0.0% | Target 0.319 · CE 6.24 · Genuine 4.49 · WER 0.0% |
| Seed 1702 J’ai trouvé la petite boîte. Elle était près de la vieille fenêtre. Nous pouvons regarder ensemble ce qu’elle contient. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] J’ai trouvé la petite boîte. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Elle était près de la vieille fenêtre. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Nous pouvons regarder ensemble ce qu’elle contient. | Target 0.312 · CE 6.19 · Genuine 4.77 · WER 0.0% | Target 0.298 · CE 6.23 · Genuine 4.43 · WER 0.0% | Target 0.337 · CE 6.17 · Genuine 4.66 · WER 0.0% | Target 0.295 · CE 6.24 · Genuine 4.42 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.387 ± 0.015 | 6.118 ± 0.044 | 4.256 ± 0.139 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.392 ± 0.006 | 6.124 ± 0.126 | 4.296 ± 0.068 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.372 ± 0.015 | 6.076 ± 0.062 | 4.267 ± 0.179 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.397 ± 0.007 | 6.162 ± 0.088 | 4.266 ± 0.072 | 0.00% ± 0.00 |
| Prompt / seed | Merged base · no new LoRA | Rank32 × 0.5 | Rank32 × 0.75 | Rank32 × 1.0 |
|---|---|---|---|---|
| Seed 1701 De deur staat nog open. Ik heb je jas op de stoel gelegd. Zeg maar wanneer je klaar bent. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] De deur staat nog open. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Ik heb je jas op de stoel gelegd. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Zeg maar wanneer je klaar bent. | Target 0.398 · CE 6.18 · Genuine 4.26 · WER 0.0% | Target 0.396 · CE 6.21 · Genuine 4.34 · WER 0.0% | Target 0.369 · CE 5.99 · Genuine 4.14 · WER 0.0% | Target 0.402 · CE 6.15 · Genuine 4.26 · WER 0.0% |
| Seed 1702 De deur staat nog open. Ik heb je jas op de stoel gelegd. Zeg maar wanneer je klaar bent. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] De deur staat nog open. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Ik heb je jas op de stoel gelegd. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Zeg maar wanneer je klaar bent. | Target 0.393 · CE 6.10 · Genuine 4.27 · WER 0.0% | Target 0.398 · CE 5.94 · Genuine 4.21 · WER 0.0% | Target 0.391 · CE 6.12 · Genuine 4.52 · WER 0.0% | Target 0.388 · CE 6.06 · Genuine 4.37 · WER 0.0% |
| Seed 1701 Ik heb het kleine doosje gevonden. Het stond naast het oude raam. We kunnen er samen in kijken. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Ik heb het kleine doosje gevonden. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Het stond naast het oude raam. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] We kunnen er samen in kijken. | Target 0.366 · CE 6.08 · Genuine 4.08 · WER 0.0% | Target 0.386 · CE 6.13 · Genuine 4.36 · WER 0.0% | Target 0.355 · CE 6.07 · Genuine 4.14 · WER 0.0% | Target 0.398 · CE 6.16 · Genuine 4.23 · WER 0.0% |
| Seed 1702 Ik heb het kleine doosje gevonden. Het stond naast het oude raam. We kunnen er samen in kijken. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Ik heb het kleine doosje gevonden. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Het stond naast het oude raam. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] We kunnen er samen in kijken. | Target 0.392 · CE 6.11 · Genuine 4.42 · WER 0.0% | Target 0.389 · CE 6.21 · Genuine 4.27 · WER 0.0% | Target 0.374 · CE 6.13 · Genuine 4.27 · WER 0.0% | Target 0.401 · CE 6.28 · Genuine 4.21 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.329 ± 0.028 | 5.998 ± 0.168 | 4.231 ± 0.369 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.352 ± 0.038 | 6.082 ± 0.146 | 4.325 ± 0.291 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.332 ± 0.041 | 6.106 ± 0.135 | 4.454 ± 0.277 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.338 ± 0.044 | 6.076 ± 0.170 | 4.220 ± 0.466 | 0.00% ± 0.00 |
| Prompt / seed | Merged base · no new LoRA | Rank32 × 0.5 | Rank32 × 0.75 | Rank32 × 1.0 |
|---|---|---|---|---|
| Seed 1701 La puerta sigue abierta. He dejado tu abrigo sobre la silla. Dime cuándo estás listo. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] La puerta sigue abierta. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] He dejado tu abrigo sobre la silla. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Dime cuándo estás listo. | Target 0.355 · CE 6.05 · Genuine 4.18 · WER 0.0% | Target 0.391 · CE 5.88 · Genuine 4.05 · WER 0.0% | Target 0.355 · CE 6.05 · Genuine 4.23 · WER 0.0% | Target 0.378 · CE 6.05 · Genuine 3.95 · WER 0.0% |
| Seed 1702 La puerta sigue abierta. He dejado tu abrigo sobre la silla. Dime cuándo estás listo. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] La puerta sigue abierta. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] He dejado tu abrigo sobre la silla. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Dime cuándo estás listo. | Target 0.352 · CE 5.75 · Genuine 3.77 · WER 0.0% | Target 0.377 · CE 6.07 · Genuine 4.10 · WER 0.0% | Target 0.374 · CE 6.04 · Genuine 4.34 · WER 0.0% | Target 0.373 · CE 5.85 · Genuine 3.70 · WER 0.0% |
| Seed 1701 He encontrado la cajita. Estaba junto a la ventana vieja. Podemos mirar juntos lo que hay dentro. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] He encontrado la cajita. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Estaba junto a la ventana vieja. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Podemos mirar juntos lo que hay dentro. | Target 0.300 · CE 6.10 · Genuine 4.32 · WER 0.0% | Target 0.309 · CE 6.19 · Genuine 4.56 · WER 0.0% | Target 0.316 · CE 6.03 · Genuine 4.86 · WER 0.0% | Target 0.306 · CE 6.21 · Genuine 4.62 · WER 0.0% |
| Seed 1702 He encontrado la cajita. Estaba junto a la ventana vieja. Podemos mirar juntos lo que hay dentro. Exact GENERAL / SCRIPT promptGENERAL: One adult speaker in a clean recording. Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] He encontrado la cajita. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.8 seconds duration] Estaba junto a la ventana vieja. (Intense Contempt. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Podemos mirar juntos lo que hay dentro. | Target 0.311 · CE 6.09 · Genuine 4.66 · WER 0.0% | Target 0.331 · CE 6.19 · Genuine 4.60 · WER 0.0% | Target 0.283 · CE 6.31 · Genuine 4.38 · WER 0.0% | Target 0.294 · CE 6.21 · Genuine 4.61 · WER 0.0% |