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 2.0.
| Held-out condition | n | Target z↑ | Raw head | CE↑ | Genuine↑ | Voice↑ | WER↓ | Valid | QA |
|---|---|---|---|---|---|---|---|---|---|
| standard | 50 | 0.649 | 0.558 | 6.269 | 4.342 | 0.552 | 0.010 | 1.000 | pass |
| manual | 50 | 1.213 | 0.851 | 6.253 | 4.151 | 0.515 | 0.017 | 0.960 | pass |
| adapter | 50 | 1.640 | 1.073 | 6.250 | 4.041 | 0.554 | 0.007 | 1.000 | pass |
| matched_no_adapter | 50 | 1.553 | 1.028 | 6.232 | 4.059 | 0.545 | 0.005 | 1.000 | pass |
Incremental adapter gain versus same-CFG control: 0.087 z;95% cluster interval [-0.056757816436399086, 0.29349608038908237]. Versus best manual-only: 0.427 z. Manual-only versus standard: 0.564 z.
| Scale | CFG | Target z | CE | Genuine | WER | Provisional QA |
|---|---|---|---|---|---|---|
| 0.0 | 1.0 | 1.288 | 6.191 | 4.186 | 0.003 | True |
| 0.5 | 1.0 | 1.139 | 6.191 | 4.189 | 0.006 | False |
| 0.75 | 1.0 | 1.294 | 6.167 | 4.199 | 0.003 | True |
| 1.0 | 1.0 | 1.190 | 6.220 | 4.242 | 0.009 | True |
| 0.0 | 1.25 | 1.370 | 6.165 | 4.116 | 0.009 | True |
| 0.5 | 1.25 | 1.200 | 6.191 | 4.284 | 0.000 | True |
| 0.75 | 1.25 | 1.475 | 6.209 | 4.158 | 0.000 | True |
| 1.0 | 1.25 | 1.307 | 6.214 | 4.251 | 0.003 | True |
| 0.0 | 1.5 | 1.399 | 6.197 | 4.178 | 0.007 | True |
| 0.5 | 1.5 | 1.384 | 6.203 | 4.191 | 0.006 | False |
| 0.75 | 1.5 | 1.400 | 6.209 | 4.146 | 0.005 | True |
| 1.0 | 1.5 | 1.432 | 6.171 | 4.110 | 0.000 | True |
| 0.0 | 2.0 | 1.571 | 6.213 | 4.154 | 0.000 | True |
| 0.5 | 2.0 | 1.653 | 6.203 | 4.162 | 0.009 | True |
| 0.75 | 2.0 | 1.631 | 6.237 | 4.184 | 0.003 | True |
| 1.0 | 2.0 | 1.637 | 6.176 | 4.127 | 0.009 | True |
| 0.0 | 2.5 | 2.250 | 6.157 | 4.118 | 0.009 | False |
| 0.5 | 2.5 | 1.809 | 6.190 | 4.119 | 0.008 | False |
| 0.75 | 2.5 | 1.922 | 6.131 | 4.072 | 0.006 | False |
| 1.0 | 2.5 | 1.721 | 6.148 | 4.106 | 0.000 | False |
| 0.0 | 3.0 | 2.011 | 6.149 | 4.087 | 0.009 | False |
| 0.5 | 3.0 | 2.017 | 6.146 | 4.034 | 0.007 | True |
| 0.75 | 3.0 | 1.904 | 6.142 | 4.094 | 0.018 | False |
| 1.0 | 3.0 | 1.913 | 6.169 | 4.047 | 0.013 | False |
| 0.0 | 4.0 | 2.202 | 6.112 | 3.981 | 0.009 | False |
| 0.5 | 4.0 | 2.245 | 6.134 | 4.024 | 0.013 | False |
| 0.75 | 4.0 | 2.089 | 6.084 | 4.035 | 0.031 | False |
| 1.0 | 4.0 | 1.920 | 6.129 | 4.029 | 0.022 | False |
| 0.0 | 5.0 | 2.883 | 6.042 | 4.000 | 0.045 | False |
| 0.5 | 5.0 | 2.738 | 6.058 | 4.023 | 0.029 | False |
| 0.75 | 5.0 | 2.800 | 6.088 | 3.974 | 0.018 | False |
| 1.0 | 5.0 | 2.323 | 6.038 | 4.047 | 0.029 | False |
direct_manual_recipe
{
"id": "manual-b8282fc33fc9",
"general": "Focus on precise articulation and clarity, using controlled projection to make the pride feel earned and meticulous rather than boastful.",
"inline": [
"Initiate with crisp, deliberate consonants, projecting a focused sense of personal victory.",
"Sustain the mid-phrase with a clear, unwavering pitch contour that speaks to deep conviction.",
"Conclude with a precise, resonant finality, letting the volume drop slightly while maintaining tonal authority."
],
"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 Pride. 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": 11, "en": 19}. 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 | 0.773 ± 0.139 | 6.213 ± 0.123 | 4.322 ± 0.227 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 20 | 0.756 ± 0.137 | 6.203 ± 0.138 | 4.322 ± 0.200 | 1.94% ± 8.70 |
| Rank32 × 0.75 | 20 | 0.766 ± 0.130 | 6.194 ± 0.102 | 4.338 ± 0.251 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 20 | 0.734 ± 0.138 | 6.191 ± 0.160 | 4.344 ± 0.229 | 0.33% ± 1.49 |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.748 ± 0.130 | 6.293 ± 0.093 | 4.280 ± 0.357 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.756 ± 0.162 | 6.279 ± 0.091 | 4.206 ± 0.371 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.761 ± 0.122 | 6.287 ± 0.031 | 4.242 ± 0.330 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.741 ± 0.153 | 6.342 ± 0.083 | 4.310 ± 0.229 | 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.832 · CE 6.22 · Genuine 3.96 · WER 0.0% | Target 0.873 · CE 6.19 · Genuine 3.88 · WER 0.0% | Target 0.856 · CE 6.32 · Genuine 4.17 · WER 0.0% | Target 0.860 · CE 6.25 · Genuine 4.22 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.876 · CE 6.25 · Genuine 3.98 · WER 0.0% | Target 0.918 · CE 6.22 · Genuine 3.89 · WER 0.0% | Target 0.875 · CE 6.25 · Genuine 3.82 · WER 0.0% | Target 0.887 · CE 6.29 · Genuine 4.03 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.594 · CE 6.28 · Genuine 4.61 · WER 0.0% | Target 0.610 · CE 6.39 · Genuine 4.57 · WER 0.0% | Target 0.637 · CE 6.31 · Genuine 4.58 · WER 0.0% | Target 0.617 · CE 6.43 · Genuine 4.53 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.691 · CE 6.43 · Genuine 4.57 · WER 0.0% | Target 0.623 · CE 6.32 · Genuine 4.48 · WER 0.0% | Target 0.675 · CE 6.27 · Genuine 4.40 · WER 0.0% | Target 0.602 · CE 6.40 · Genuine 4.46 · WER 0.0% |
Fixed speaker reference:
| Condition | n | Target score ↑ | Content Enjoyment ↑ | Genuineness ↑ | WER ↓ |
|---|---|---|---|---|---|
| Merged base, no bucket adapter | 4 | 0.921 ± 0.094 | 6.279 ± 0.118 | 4.333 ± 0.073 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.899 ± 0.114 | 6.338 ± 0.129 | 4.372 ± 0.058 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.880 ± 0.133 | 6.251 ± 0.079 | 4.401 ± 0.030 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.898 ± 0.119 | 6.303 ± 0.174 | 4.413 ± 0.225 | 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.813 · CE 6.19 · Genuine 4.39 · WER 0.0% | Target 0.863 · CE 6.33 · Genuine 4.33 · WER 0.0% | Target 0.802 · CE 6.16 · Genuine 4.37 · WER 0.0% | Target 0.835 · CE 6.10 · Genuine 4.36 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.871 · CE 6.17 · Genuine 4.38 · WER 0.0% | Target 0.754 · CE 6.16 · Genuine 4.39 · WER 0.0% | Target 0.734 · CE 6.22 · Genuine 4.42 · WER 0.0% | Target 0.763 · CE 6.22 · Genuine 4.60 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Wir können gemeinsam hineinschauen. | Target 1.000 · CE 6.33 · Genuine 4.23 · WER 0.0% | Target 0.986 · CE 6.43 · Genuine 4.44 · WER 0.0% | Target 1.000 · CE 6.28 · Genuine 4.38 · WER 0.0% | Target 0.999 · CE 6.49 · Genuine 4.58 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Wir können gemeinsam hineinschauen. | Target 1.000 · CE 6.42 · Genuine 4.33 · WER 0.0% | Target 0.994 · CE 6.44 · Genuine 4.32 · WER 0.0% | Target 0.986 · CE 6.34 · Genuine 4.43 · WER 0.0% | Target 0.996 · CE 6.40 · Genuine 4.12 · 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 | 0.823 ± 0.081 | 6.229 ± 0.115 | 4.462 ± 0.193 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.785 ± 0.062 | 6.197 ± 0.129 | 4.388 ± 0.165 | 9.72% ± 19.44 |
| Rank32 × 0.75 | 4 | 0.820 ± 0.022 | 6.146 ± 0.043 | 4.485 ± 0.225 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.755 ± 0.070 | 6.130 ± 0.160 | 4.496 ± 0.290 | 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.744 · CE 6.08 · Genuine 4.39 · WER 0.0% | Target 0.801 · CE 6.30 · Genuine 4.38 · WER 38.9% | Target 0.838 · CE 6.13 · Genuine 4.37 · WER 0.0% | Target 0.757 · CE 6.31 · Genuine 4.62 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.816 · CE 6.25 · Genuine 4.28 · WER 0.0% | Target 0.724 · CE 6.09 · Genuine 4.18 · WER 0.0% | Target 0.789 · CE 6.09 · Genuine 4.23 · WER 0.0% | Target 0.809 · CE 6.21 · Genuine 4.06 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.935 · CE 6.36 · Genuine 4.44 · WER 0.0% | Target 0.865 · CE 6.32 · Genuine 4.58 · WER 0.0% | Target 0.833 · CE 6.16 · Genuine 4.72 · WER 0.0% | Target 0.801 · CE 6.06 · Genuine 4.65 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.798 · CE 6.23 · Genuine 4.73 · WER 0.0% | Target 0.749 · CE 6.09 · Genuine 4.41 · WER 0.0% | Target 0.820 · CE 6.20 · Genuine 4.62 · WER 0.0% | Target 0.656 · CE 5.94 · Genuine 4.66 · 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 | 0.621 ± 0.105 | 6.170 ± 0.079 | 4.272 ± 0.129 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.581 ± 0.061 | 6.150 ± 0.059 | 4.349 ± 0.156 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.631 ± 0.122 | 6.144 ± 0.090 | 4.322 ± 0.207 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.578 ± 0.060 | 6.078 ± 0.093 | 4.266 ± 0.084 | 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.658 · CE 6.19 · Genuine 4.16 · WER 0.0% | Target 0.566 · CE 6.19 · Genuine 4.12 · WER 0.0% | Target 0.770 · CE 6.15 · Genuine 4.14 · WER 0.0% | Target 0.569 · CE 6.17 · Genuine 4.31 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.590 · CE 6.07 · Genuine 4.29 · WER 0.0% | Target 0.663 · CE 6.06 · Genuine 4.39 · WER 0.0% | Target 0.474 · CE 6.01 · Genuine 4.58 · WER 0.0% | Target 0.517 · CE 5.95 · Genuine 4.28 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.740 · CE 6.26 · Genuine 4.19 · WER 0.0% | Target 0.517 · CE 6.17 · Genuine 4.43 · WER 0.0% | Target 0.656 · CE 6.21 · Genuine 4.39 · WER 0.0% | Target 0.661 · CE 6.13 · Genuine 4.14 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. 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 Pride. 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 Pride. 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 0.494 · CE 6.16 · Genuine 4.44 · WER 0.0% | Target 0.580 · CE 6.17 · Genuine 4.46 · WER 0.0% | Target 0.624 · CE 6.20 · Genuine 4.17 · WER 0.0% | Target 0.564 · CE 6.07 · Genuine 4.33 · 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 | 0.750 ± 0.122 | 6.096 ± 0.134 | 4.261 ± 0.322 | 0.00% ± 0.00 |
| Rank32 × 0.5 | 4 | 0.757 ± 0.062 | 6.048 ± 0.088 | 4.294 ± 0.180 | 0.00% ± 0.00 |
| Rank32 × 0.75 | 4 | 0.738 ± 0.111 | 6.142 ± 0.150 | 4.241 ± 0.368 | 0.00% ± 0.00 |
| Rank32 × 1.0 | 4 | 0.695 ± 0.065 | 6.100 ± 0.098 | 4.236 ± 0.273 | 1.67% ± 3.33 |
| 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] La puerta sigue abierta. (Intense Pride. 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Dime cuándo estás listo. | Target 0.868 · CE 6.20 · Genuine 4.17 · WER 0.0% | Target 0.724 · CE 5.97 · Genuine 4.09 · WER 0.0% | Target 0.817 · CE 6.04 · Genuine 3.94 · WER 0.0% | Target 0.773 · CE 6.05 · Genuine 4.05 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] La puerta sigue abierta. (Intense Pride. 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] Dime cuándo estás listo. | Target 0.761 · CE 5.95 · Genuine 3.85 · WER 0.0% | Target 0.851 · CE 6.06 · Genuine 4.19 · WER 0.0% | Target 0.841 · CE 6.02 · Genuine 4.02 · WER 0.0% | Target 0.722 · CE 6.00 · Genuine 3.96 · WER 6.7% |
| 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] He encontrado la cajita. (Intense Pride. 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 Pride. 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 0.580 · CE 6.01 · Genuine 4.47 · WER 0.0% | Target 0.728 · CE 6.00 · Genuine 4.47 · WER 0.0% | Target 0.687 · CE 6.16 · Genuine 4.24 · WER 0.0% | Target 0.655 · CE 6.22 · Genuine 4.46 · 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 Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word. SCRIPT: (Intense Pride. Make the requested delivery clearly audible, not merely implied by the words. Preserve every word.) [3.2 seconds duration] He encontrado la cajita. (Intense Pride. 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 Pride. 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 0.791 · CE 6.22 · Genuine 4.56 · WER 0.0% | Target 0.726 · CE 6.17 · Genuine 4.42 · WER 0.0% | Target 0.605 · CE 6.35 · Genuine 4.76 · WER 0.0% | Target 0.630 · CE 6.13 · Genuine 4.48 · WER 0.0% |