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 4.0.
| Held-out condition | n | Target z↑ | Raw head | CE↑ | Genuine↑ | Voice↑ | WER↓ | Valid | QA |
|---|---|---|---|---|---|---|---|---|---|
| standard | 50 | -0.366 | 0.019 | 6.250 | 4.395 | 0.578 | 0.013 | 1.000 | pass |
| manual | 50 | -0.328 | 0.039 | 5.977 | 4.381 | 0.600 | 0.025 | 0.980 | pass |
| adapter | 50 | -0.338 | 0.034 | 5.887 | 4.318 | 0.598 | 0.026 | 1.000 | pass |
| matched_no_adapter | 50 | -0.327 | 0.040 | 5.863 | 4.288 | 0.558 | 0.028 | 1.000 | pass |
Incremental adapter gain versus same-CFG control: -0.010 z;95% cluster interval [-0.02437500376779208, 0.004101565001494032]. Versus best manual-only: -0.009 z. Manual-only versus standard: 0.038 z.
| Scale | CFG | Target z | CE | Genuine | WER | Provisional QA |
|---|---|---|---|---|---|---|
| 0.0 | 1.0 | -0.367 | 6.092 | 4.470 | 0.011 | True |
| 0.5 | 1.0 | -0.366 | 6.153 | 4.501 | 0.000 | True |
| 0.75 | 1.0 | -0.362 | 6.151 | 4.463 | 0.003 | True |
| 1.0 | 1.0 | -0.358 | 6.162 | 4.469 | 0.011 | True |
| 0.0 | 1.25 | -0.368 | 6.165 | 4.505 | 0.003 | True |
| 0.5 | 1.25 | -0.370 | 6.137 | 4.529 | 0.003 | True |
| 0.75 | 1.25 | -0.373 | 6.134 | 4.516 | 0.000 | True |
| 1.0 | 1.25 | -0.363 | 6.136 | 4.495 | 0.008 | True |
| 0.0 | 1.5 | -0.366 | 6.127 | 4.481 | 0.005 | True |
| 0.5 | 1.5 | -0.369 | 6.105 | 4.542 | 0.000 | True |
| 0.75 | 1.5 | -0.370 | 6.123 | 4.473 | 0.003 | True |
| 1.0 | 1.5 | -0.352 | 6.121 | 4.554 | 0.000 | True |
| 0.0 | 2.0 | -0.354 | 6.108 | 4.560 | 0.014 | True |
| 0.5 | 2.0 | -0.356 | 6.100 | 4.557 | 0.006 | True |
| 0.75 | 2.0 | -0.360 | 6.039 | 4.526 | 0.000 | True |
| 1.0 | 2.0 | -0.363 | 6.076 | 4.604 | 0.000 | True |
| 0.0 | 2.5 | -0.365 | 6.047 | 4.597 | 0.006 | True |
| 0.5 | 2.5 | -0.360 | 6.066 | 4.576 | 0.005 | True |
| 0.75 | 2.5 | -0.360 | 6.104 | 4.608 | 0.006 | True |
| 1.0 | 2.5 | -0.352 | 6.050 | 4.598 | 0.003 | True |
| 0.0 | 3.0 | -0.355 | 5.926 | 4.580 | 0.016 | True |
| 0.5 | 3.0 | -0.350 | 6.011 | 4.573 | 0.003 | True |
| 0.75 | 3.0 | -0.347 | 6.051 | 4.622 | 0.003 | True |
| 1.0 | 3.0 | -0.349 | 5.980 | 4.576 | 0.005 | True |
| 0.0 | 4.0 | -0.350 | 5.886 | 4.552 | 0.012 | True |
| 0.5 | 4.0 | -0.348 | 5.861 | 4.504 | 0.008 | True |
| 0.75 | 4.0 | -0.346 | 5.871 | 4.441 | 0.000 | True |
| 1.0 | 4.0 | -0.346 | 5.914 | 4.499 | 0.009 | True |
| 0.0 | 5.0 | -0.343 | 5.782 | 4.411 | 0.011 | False |
| 0.5 | 5.0 | -0.339 | 5.764 | 4.440 | 0.005 | False |
| 0.75 | 5.0 | -0.337 | 5.834 | 4.466 | 0.009 | True |
| 1.0 | 5.0 | -0.343 | 5.727 | 4.462 | 0.025 | True |
direct_manual_recipe
{
"id": "manual-48b6c1c43c18",
"general": "Treat the vocal tract as constricted; the shame manifests as a physical tightening around the mouth and throat.",
"inline": [
"Begin with a constricted resonance, as if the vocal cords are physically strained before speaking.",
"Maintain a very tight, pressurized quality throughout, preventing any natural vocal release or expansion.",
"Conclude with a sudden, sharp intake of breath that punctuates the silence immediately after the final utterance."
],
"cfg": 3,
"recipe": "P05",
"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 Shame. 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": 12, "fr": 5, "de": 8, "es": 6}. 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 | 0.356 ± 0.016 | 6.191 ± 0.127 | 4.377 ± 0.231 | 0.33% ± 1.49 |
| Rank32 × 0.5 | 20 | 0.359 ± 0.014 | 6.183 ± 0.167 | 4.329 ± 0.220 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 20 | 0.357 ± 0.014 | 6.207 ± 0.123 | 4.362 ± 0.211 | 0.53% ± 2.35 |
| Rank32 × 1.0 | 20 | 0.356 ± 0.015 | 6.168 ± 0.146 | 4.388 ± 0.188 | 0.28% ± 1.24 |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.345 ± 0.011 | 6.275 ± 0.129 | 4.208 ± 0.303 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.346 ± 0.013 | 6.285 ± 0.122 | 4.150 ± 0.331 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.345 ± 0.008 | 6.289 ± 0.142 | 4.313 ± 0.360 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.340 ± 0.006 | 6.320 ± 0.076 | 4.242 ± 0.153 | 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.359 · CE 6.22 · Genuine 3.93 · WER 0.0% | Target 0.358 · CE 6.16 · Genuine 4.04 · WER 0.0% | Target 0.348 · CE 6.25 · Genuine 4.11 · WER 0.0% | Target 0.347 · CE 6.28 · Genuine 4.05 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.349 · CE 6.12 · Genuine 4.04 · WER 0.0% | Target 0.352 · CE 6.21 · Genuine 3.74 · WER 0.0% | Target 0.355 · CE 6.11 · Genuine 3.93 · WER 0.0% | Target 0.344 · CE 6.35 · Genuine 4.20 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.339 · CE 6.36 · Genuine 4.24 · WER 0.0% | Target 0.329 · CE 6.41 · Genuine 4.48 · WER 0.0% | Target 0.340 · CE 6.35 · Genuine 4.72 · WER 0.0% | Target 0.335 · CE 6.41 · Genuine 4.32 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.334 · CE 6.40 · Genuine 4.62 · WER 0.0% | Target 0.344 · CE 6.36 · Genuine 4.34 · WER 0.0% | Target 0.337 · CE 6.44 · Genuine 4.49 · WER 0.0% | Target 0.334 · CE 6.24 · Genuine 4.40 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.363 ± 0.006 | 6.267 ± 0.096 | 4.405 ± 0.179 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.371 ± 0.015 | 6.344 ± 0.149 | 4.407 ± 0.163 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.369 ± 0.017 | 6.307 ± 0.092 | 4.369 ± 0.213 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.372 ± 0.011 | 6.283 ± 0.121 | 4.405 ± 0.125 | 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.360 · CE 6.20 · Genuine 4.41 · WER 0.0% | Target 0.382 · CE 6.32 · Genuine 4.58 · WER 0.0% | Target 0.385 · CE 6.24 · Genuine 4.65 · WER 0.0% | Target 0.380 · CE 6.17 · Genuine 4.43 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.371 · CE 6.17 · Genuine 4.66 · WER 0.0% | Target 0.386 · CE 6.14 · Genuine 4.22 · WER 0.0% | Target 0.383 · CE 6.22 · Genuine 4.41 · WER 0.0% | Target 0.383 · CE 6.19 · Genuine 4.57 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Wir können gemeinsam hineinschauen. | Target 0.363 · CE 6.37 · Genuine 4.29 · WER 0.0% | Target 0.363 · CE 6.44 · Genuine 4.49 · WER 0.0% | Target 0.356 · CE 6.41 · Genuine 4.14 · WER 0.0% | Target 0.366 · CE 6.42 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Wir können gemeinsam hineinschauen. | Target 0.357 · CE 6.32 · Genuine 4.27 · WER 0.0% | Target 0.353 · CE 6.48 · Genuine 4.33 · WER 0.0% | Target 0.354 · CE 6.36 · Genuine 4.28 · WER 0.0% | Target 0.360 · CE 6.35 · Genuine 4.29 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.343 ± 0.010 | 6.187 ± 0.110 | 4.492 ± 0.151 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.359 ± 0.004 | 6.070 ± 0.120 | 4.441 ± 0.121 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.355 ± 0.002 | 6.152 ± 0.094 | 4.443 ± 0.118 | 2.63% ± 5.26 |
| Rank32 × 1.0 | 4 | 0.353 ± 0.010 | 6.096 ± 0.042 | 4.402 ± 0.126 | 1.39% ± 2.78 |
| 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.327 · CE 6.21 · Genuine 4.63 · WER 0.0% | Target 0.361 · CE 6.09 · Genuine 4.38 · WER 0.0% | Target 0.355 · CE 6.08 · Genuine 4.48 · WER 0.0% | Target 0.338 · CE 6.05 · Genuine 4.53 · WER 5.6% |
| 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.347 · CE 6.08 · Genuine 4.46 · WER 0.0% | Target 0.361 · CE 6.22 · Genuine 4.36 · WER 0.0% | Target 0.355 · CE 6.06 · Genuine 4.31 · WER 0.0% | Target 0.359 · CE 6.12 · Genuine 4.24 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.347 · CE 6.33 · Genuine 4.58 · WER 0.0% | Target 0.362 · CE 5.94 · Genuine 4.62 · WER 0.0% | Target 0.358 · CE 6.23 · Genuine 4.58 · WER 0.0% | Target 0.356 · CE 6.08 · Genuine 4.46 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.350 · CE 6.13 · Genuine 4.29 · WER 0.0% | Target 0.354 · CE 6.03 · Genuine 4.41 · WER 0.0% | Target 0.352 · CE 6.24 · Genuine 4.40 · WER 10.5% | Target 0.359 · CE 6.14 · Genuine 4.38 · WER 0.0% |
No accepted bucket-training example in this language: transfer evaluation only.
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.374 ± 0.013 | 6.128 ± 0.046 | 4.322 ± 0.142 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.368 ± 0.009 | 6.124 ± 0.117 | 4.305 ± 0.058 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.369 ± 0.007 | 6.164 ± 0.054 | 4.324 ± 0.082 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.364 ± 0.016 | 6.086 ± 0.150 | 4.313 ± 0.073 | 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.367 · CE 6.13 · Genuine 4.34 · WER 0.0% | Target 0.380 · CE 6.21 · Genuine 4.36 · WER 0.0% | Target 0.369 · CE 6.21 · Genuine 4.33 · WER 0.0% | Target 0.373 · CE 6.11 · Genuine 4.37 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.374 · CE 6.19 · Genuine 4.33 · WER 0.0% | Target 0.364 · CE 6.18 · Genuine 4.23 · WER 0.0% | Target 0.374 · CE 6.17 · Genuine 4.24 · WER 0.0% | Target 0.374 · CE 6.14 · Genuine 4.24 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.393 · CE 6.09 · Genuine 4.14 · WER 0.0% | Target 0.368 · CE 5.95 · Genuine 4.29 · WER 0.0% | Target 0.374 · CE 6.20 · Genuine 4.29 · WER 0.0% | Target 0.367 · CE 6.22 · Genuine 4.26 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. 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 Shame. 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 Shame. 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 0.363 · CE 6.10 · Genuine 4.48 · WER 0.0% | Target 0.360 · CE 6.16 · Genuine 4.34 · WER 0.0% | Target 0.359 · CE 6.09 · Genuine 4.43 · WER 0.0% | Target 0.341 · CE 5.87 · Genuine 4.38 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.352 ± 0.014 | 6.096 ± 0.166 | 4.459 ± 0.318 | 1.67% ± 3.33 |
| Rank32 × 0.5 | 4 | 0.350 ± 0.013 | 6.090 ± 0.171 | 4.341 ± 0.288 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.348 ± 0.008 | 6.124 ± 0.128 | 4.363 ± 0.270 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.349 ± 0.010 | 6.056 ± 0.094 | 4.576 ± 0.281 | 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] La puerta sigue abierta. (Intense Shame. 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Dime cuándo estás listo. | Target 0.357 · CE 6.14 · Genuine 4.19 · WER 0.0% | Target 0.363 · CE 5.93 · Genuine 4.04 · WER 0.0% | Target 0.357 · CE 6.22 · Genuine 4.17 · WER 0.0% | Target 0.362 · CE 5.98 · Genuine 4.24 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] La puerta sigue abierta. (Intense Shame. 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Dime cuándo estás listo. | Target 0.369 · CE 5.87 · Genuine 4.17 · WER 6.7% | Target 0.359 · CE 5.96 · Genuine 4.17 · WER 0.0% | Target 0.352 · CE 6.18 · Genuine 4.09 · WER 0.0% | Target 0.352 · CE 6.00 · Genuine 4.46 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] He encontrado la cajita. (Intense Shame. 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 Shame. 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 0.337 · CE 6.27 · Genuine 4.76 · WER 0.0% | Target 0.335 · CE 6.25 · Genuine 4.67 · WER 0.0% | Target 0.340 · CE 6.16 · Genuine 4.61 · WER 0.0% | Target 0.341 · CE 6.05 · Genuine 4.73 · 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 Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Shame. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] He encontrado la cajita. (Intense Shame. 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 Shame. 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 0.347 · CE 6.10 · Genuine 4.71 · WER 0.0% | Target 0.342 · CE 6.22 · Genuine 4.48 · WER 0.0% | Target 0.343 · CE 5.94 · Genuine 4.58 · WER 0.0% | Target 0.342 · CE 6.19 · Genuine 4.87 · WER 0.0% |