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: Use manual-only configuration; no clear qualifying incremental adapter gain. Bucket scale 0.0, CFG 5.0.
| Held-out condition | n | Target z↑ | Raw head | CE↑ | Genuine↑ | Voice↑ | WER↓ | Valid | QA |
|---|---|---|---|---|---|---|---|---|---|
| standard | 50 | -0.020 | 0.356 | 6.244 | 4.401 | 0.594 | 0.007 | 1.000 | pass |
| manual | 50 | 0.487 | 0.710 | 6.222 | 4.417 | 0.555 | 0.065 | 0.920 | pass |
| adapter | 50 | 0.310 | 0.586 | 6.254 | 4.402 | 0.562 | 0.048 | 0.920 | pass |
| matched_no_adapter | 50 | 0.321 | 0.594 | 6.251 | 4.393 | 0.561 | 0.038 | 0.940 | pass |
Incremental adapter gain versus same-CFG control: -0.011 z;95% cluster interval [-0.0899609414967111, 0.08083496699845569]. Versus best manual-only: -0.177 z. Manual-only versus standard: 0.508 z.
| Scale | CFG | Target z | CE | Genuine | WER | Provisional QA |
|---|---|---|---|---|---|---|
| 0.0 | 1.0 | 0.379 | 6.213 | 4.372 | 0.024 | True |
| 0.5 | 1.0 | 0.429 | 6.203 | 4.340 | 0.003 | False |
| 0.75 | 1.0 | 0.362 | 6.185 | 4.304 | 0.003 | False |
| 1.0 | 1.0 | 0.473 | 6.207 | 4.285 | 0.006 | True |
| 0.0 | 1.25 | 0.373 | 6.216 | 4.330 | 0.000 | False |
| 0.5 | 1.25 | 0.454 | 6.133 | 4.268 | 0.013 | True |
| 0.75 | 1.25 | 0.396 | 6.187 | 4.319 | 0.000 | True |
| 1.0 | 1.25 | 0.365 | 6.230 | 4.347 | 0.005 | True |
| 0.0 | 1.5 | 0.588 | 6.219 | 4.291 | 0.005 | True |
| 0.5 | 1.5 | 0.489 | 6.166 | 4.357 | 0.003 | False |
| 0.75 | 1.5 | 0.433 | 6.198 | 4.301 | 0.032 | True |
| 1.0 | 1.5 | 0.418 | 6.180 | 4.294 | 0.009 | False |
| 0.0 | 2.0 | 0.584 | 6.191 | 4.274 | 0.003 | True |
| 0.5 | 2.0 | 0.545 | 6.167 | 4.225 | 0.003 | False |
| 0.75 | 2.0 | 0.566 | 6.177 | 4.297 | 0.008 | True |
| 1.0 | 2.0 | 0.557 | 6.197 | 4.306 | 0.014 | False |
| 0.0 | 2.5 | 0.542 | 6.196 | 4.306 | 0.013 | True |
| 0.5 | 2.5 | 0.577 | 6.192 | 4.316 | 0.040 | False |
| 0.75 | 2.5 | 0.531 | 6.179 | 4.352 | 0.003 | True |
| 1.0 | 2.5 | 0.569 | 6.199 | 4.290 | 0.006 | True |
| 0.0 | 3.0 | 0.609 | 6.194 | 4.268 | 0.025 | False |
| 0.5 | 3.0 | 0.756 | 6.171 | 4.275 | 0.022 | False |
| 0.75 | 3.0 | 0.778 | 6.152 | 4.287 | 0.019 | True |
| 1.0 | 3.0 | 0.705 | 6.197 | 4.284 | 0.028 | False |
| 0.0 | 4.0 | 0.766 | 6.176 | 4.249 | 0.072 | False |
| 0.5 | 4.0 | 0.796 | 6.126 | 4.190 | 0.049 | True |
| 0.75 | 4.0 | 0.822 | 6.099 | 4.258 | 0.072 | False |
| 1.0 | 4.0 | 0.775 | 6.135 | 4.229 | 0.061 | False |
| 0.0 | 5.0 | 0.849 | 6.181 | 4.265 | 0.051 | True |
| 0.5 | 5.0 | 0.888 | 6.112 | 4.202 | 0.102 | False |
| 0.75 | 5.0 | 0.942 | 6.071 | 4.154 | 0.033 | False |
| 1.0 | 5.0 | 0.975 | 6.056 | 4.225 | 0.037 | False |
direct_manual_recipe
{
"id": "manual-c17f294b7963",
"general": "Shift the resonance forward slightly, giving the relief a bright, almost stunned quality without becoming overly bright.",
"inline": [
"Project the sound from a relaxed throat space",
"Introduce a gentle upward inflection on the key words",
"Ensure the final word carries a lingering, warm resonance"
],
"cfg": 3,
"recipe": "P01",
"strategy": "local-Gemma-mutation",
"timing_syntax": "canonical",
"canonical_labels": true,
"neutral_event": "keep",
"parent": "cfg-3",
"origin": "exact_frozen_manual_recipe"
}Exact machine-readable comparison · Agent tool recommendations
Intense Relief. 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: {"en": 29, "fr": 12, "de": 17, "es": 9, "nl": 2}. 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 | 0.527 ± 0.145 | 6.188 ± 0.155 | 4.432 ± 0.170 | 0.54% ± 1.67 |
| Rank32 × 0.5 | 20 | 0.567 ± 0.170 | 6.177 ± 0.133 | 4.359 ± 0.201 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 20 | 0.558 ± 0.174 | 6.195 ± 0.166 | 4.354 ± 0.261 | 2.37% ± 9.43 |
| Rank32 × 1.0 | 20 | 0.518 ± 0.142 | 6.131 ± 0.159 | 4.311 ± 0.222 | 0.26% ± 1.18 |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.612 ± 0.115 | 6.338 ± 0.065 | 4.437 ± 0.170 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.713 ± 0.130 | 6.301 ± 0.099 | 4.259 ± 0.227 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.688 ± 0.183 | 6.259 ± 0.224 | 4.219 ± 0.439 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.654 ± 0.137 | 6.239 ± 0.122 | 4.139 ± 0.388 | 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.488 · CE 6.26 · Genuine 4.39 · WER 0.0% | Target 0.545 · CE 6.21 · Genuine 4.16 · WER 0.0% | Target 0.499 · CE 6.13 · Genuine 3.89 · WER 0.0% | Target 0.529 · CE 6.13 · Genuine 3.77 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.569 · CE 6.31 · Genuine 4.29 · WER 0.0% | Target 0.679 · CE 6.23 · Genuine 4.00 · WER 0.0% | Target 0.563 · CE 6.01 · Genuine 3.79 · WER 0.0% | Target 0.545 · CE 6.13 · Genuine 3.84 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.629 · CE 6.39 · Genuine 4.68 · WER 0.0% | Target 0.792 · CE 6.41 · Genuine 4.36 · WER 0.0% | Target 0.852 · CE 6.44 · Genuine 4.55 · WER 0.0% | Target 0.742 · CE 6.36 · Genuine 4.55 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.761 · CE 6.40 · Genuine 4.38 · WER 0.0% | Target 0.838 · CE 6.36 · Genuine 4.52 · WER 0.0% | Target 0.838 · CE 6.45 · Genuine 4.65 · WER 0.0% | Target 0.798 · CE 6.33 · Genuine 4.39 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.471 ± 0.082 | 6.307 ± 0.133 | 4.300 ± 0.185 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.508 ± 0.108 | 6.277 ± 0.144 | 4.517 ± 0.227 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.478 ± 0.092 | 6.332 ± 0.150 | 4.482 ± 0.201 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.482 ± 0.112 | 6.322 ± 0.076 | 4.376 ± 0.093 | 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.392 · CE 6.26 · Genuine 4.41 · WER 0.0% | Target 0.424 · CE 6.09 · Genuine 4.61 · WER 0.0% | Target 0.415 · CE 6.17 · Genuine 4.75 · WER 0.0% | Target 0.408 · CE 6.31 · Genuine 4.37 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.415 · CE 6.14 · Genuine 4.46 · WER 0.0% | Target 0.432 · CE 6.26 · Genuine 4.80 · WER 0.0% | Target 0.404 · CE 6.27 · Genuine 4.53 · WER 0.0% | Target 0.389 · CE 6.22 · Genuine 4.29 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Wir können gemeinsam hineinschauen. | Target 0.568 · CE 6.38 · Genuine 4.29 · WER 0.0% | Target 0.657 · CE 6.33 · Genuine 4.33 · WER 0.0% | Target 0.605 · CE 6.52 · Genuine 4.31 · WER 0.0% | Target 0.636 · CE 6.39 · Genuine 4.33 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Wir können gemeinsam hineinschauen. | Target 0.507 · CE 6.44 · Genuine 4.04 · WER 0.0% | Target 0.519 · CE 6.43 · Genuine 4.34 · WER 0.0% | Target 0.487 · CE 6.38 · Genuine 4.34 · WER 0.0% | Target 0.496 · CE 6.37 · Genuine 4.51 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.615 ± 0.191 | 6.042 ± 0.082 | 4.451 ± 0.121 | 1.39% ± 2.78 |
| Rank32 × 0.5 | 4 | 0.688 ± 0.178 | 6.120 ± 0.114 | 4.438 ± 0.157 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.727 ± 0.176 | 6.185 ± 0.116 | 4.454 ± 0.067 | 1.32% ± 2.63 |
| Rank32 × 1.0 | 4 | 0.618 ± 0.096 | 6.029 ± 0.150 | 4.482 ± 0.181 | 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.484 · CE 5.98 · Genuine 4.36 · WER 5.6% | Target 0.587 · CE 6.10 · Genuine 4.44 · WER 0.0% | Target 0.530 · CE 6.13 · Genuine 4.43 · WER 0.0% | Target 0.477 · CE 6.01 · Genuine 4.22 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.433 · CE 6.14 · Genuine 4.37 · WER 0.0% | Target 0.489 · CE 6.19 · Genuine 4.51 · WER 0.0% | Target 0.645 · CE 6.28 · Genuine 4.51 · WER 0.0% | Target 0.645 · CE 6.21 · Genuine 4.53 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.845 · CE 6.08 · Genuine 4.62 · WER 0.0% | Target 0.825 · CE 5.97 · Genuine 4.22 · WER 0.0% | Target 0.931 · CE 6.28 · Genuine 4.51 · WER 0.0% | Target 0.664 · CE 6.05 · Genuine 4.63 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.697 · CE 5.97 · Genuine 4.46 · WER 0.0% | Target 0.852 · CE 6.22 · Genuine 4.58 · WER 0.0% | Target 0.801 · CE 6.05 · Genuine 4.37 · WER 5.3% | Target 0.686 · CE 5.84 · Genuine 4.55 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.370 ± 0.042 | 6.064 ± 0.063 | 4.355 ± 0.050 | 1.32% ± 2.63 |
| Rank32 × 0.5 | 4 | 0.353 ± 0.028 | 6.050 ± 0.082 | 4.291 ± 0.095 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.372 ± 0.021 | 6.052 ± 0.113 | 4.292 ± 0.105 | 10.53% ± 21.05 |
| Rank32 × 1.0 | 4 | 0.351 ± 0.026 | 6.072 ± 0.028 | 4.275 ± 0.058 | 1.32% ± 2.63 |
| 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.358 · CE 6.05 · Genuine 4.32 · WER 5.3% | Target 0.344 · CE 6.09 · Genuine 4.25 · WER 0.0% | Target 0.401 · CE 6.10 · Genuine 4.22 · WER 0.0% | Target 0.346 · CE 6.04 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.315 · CE 6.03 · Genuine 4.31 · WER 0.0% | Target 0.336 · CE 5.95 · Genuine 4.26 · WER 0.0% | Target 0.355 · CE 5.89 · Genuine 4.44 · WER 42.1% | Target 0.318 · CE 6.06 · Genuine 4.21 · WER 5.3% |
| 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.402 · CE 6.02 · Genuine 4.39 · WER 0.0% | Target 0.337 · CE 6.02 · Genuine 4.22 · WER 0.0% | Target 0.373 · CE 6.14 · Genuine 4.28 · WER 0.0% | Target 0.373 · CE 6.10 · Genuine 4.29 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. 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 Relief. 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 Relief. 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 0.404 · CE 6.16 · Genuine 4.41 · WER 0.0% | Target 0.394 · CE 6.14 · Genuine 4.43 · WER 0.0% | Target 0.357 · CE 6.08 · Genuine 4.22 · WER 0.0% | Target 0.369 · CE 6.09 · Genuine 4.34 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.567 ± 0.125 | 6.189 ± 0.148 | 4.617 ± 0.159 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.574 ± 0.086 | 6.138 ± 0.027 | 4.291 ± 0.228 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.527 ± 0.026 | 6.144 ± 0.120 | 4.321 ± 0.344 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.483 ± 0.087 | 5.991 ± 0.084 | 4.283 ± 0.171 | 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] La puerta sigue abierta. (Intense Relief. 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Dime cuándo estás listo. | Target 0.533 · CE 6.25 · Genuine 4.51 · WER 0.0% | Target 0.490 · CE 6.13 · Genuine 4.14 · WER 0.0% | Target 0.520 · CE 6.16 · Genuine 3.96 · WER 0.0% | Target 0.559 · CE 6.11 · Genuine 4.19 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] La puerta sigue abierta. (Intense Relief. 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Dime cuándo estás listo. | Target 0.438 · CE 6.11 · Genuine 4.46 · WER 0.0% | Target 0.530 · CE 6.17 · Genuine 4.06 · WER 0.0% | Target 0.509 · CE 6.01 · Genuine 4.09 · WER 0.0% | Target 0.397 · CE 5.92 · Genuine 4.09 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] He encontrado la cajita. (Intense Relief. 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 Relief. 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 0.737 · CE 6.36 · Genuine 4.78 · WER 0.0% | Target 0.687 · CE 6.14 · Genuine 4.53 · WER 0.0% | Target 0.566 · CE 6.30 · Genuine 4.58 · WER 0.0% | Target 0.556 · CE 5.95 · Genuine 4.42 · 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 Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Relief. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] He encontrado la cajita. (Intense Relief. 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 Relief. 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 0.562 · CE 6.03 · Genuine 4.72 · WER 0.0% | Target 0.588 · CE 6.11 · Genuine 4.43 · WER 0.0% | Target 0.512 · CE 6.10 · Genuine 4.65 · WER 0.0% | Target 0.419 · CE 5.98 · Genuine 4.43 · WER 0.0% |