Temperature-1.0 Sampling — Ranked by a Combined Reward (blend + genuineness, gated by invWER)

What am I looking at?

The same 98 voice-clone prompts as the genuineness grid (same reference speaker, same performance instruction, same paraphrased text), each sampled 10 times at temperature 1.0 from the fine-tuned MOSS-TTS 8B model. Here the 10 takes are ranked not by genuineness alone but by a combined reward that folds in perceptual quality (VoiceCLAP-small blend head, 0–10) and transcription fidelity (invWER from Parakeet-tdt-0.6b-v3) alongside genuineness.

reward_P = invWER × ( norm(blend) + norm(genu_P) )
invWER = max(0, 1 − WER) ∈ [0,1] (gates the whole reward → garbled takes score 0)
norm(·) = global min–max to [0,1] across all 962 candidates (blend, genu each normalized separately)
P ∈ { VoiceCLAP-large-v2 (3584-d), VoiceCLAP-commercial (768-d) } — shown as two separate columns

Min–max (not z-score) keeps the reward non-negative and monotone; the two genuineness terms sum with blend to [0,2], then invWER scales to [0,1]×…. Within every group we rank the 10 takes descending by reward — separately for each predictor — and show the Top-3 (green) and Bottom-2 (amber). Under each take: its combined reward and the three components (blend 0–10, genu, invWER/WER). The group's original reference clip and the spoken paraphrase are shown for context.

98 groups × 10 samples · 962 clips temperature 1.0 overall mean: blend 3.59 · invWER 0.64 · genu L 1.26/C 1.71 top-pick agreement between the two combined rewards: 71/98 = 72% combined-reward top pick differs from pure-genu top pick: large 70/98 · comm 71/98 groups with <5 variants: src_d_07(4), src_d_19(3)

Source C: blend 3.67 · invWER 0.31 · genu L 1.39/C 1.64Source B: blend 3.78 · invWER 0.86 · genu L 1.37/C 2.05German source: blend 3.94 · invWER 0.88 · genu L 1.31/C 2.02Source D: blend 5.26 · invWER 0.33 · genu L 1.42/C 1.48

All groups — interleaved order (Source C · Source B · German source · Source D, repeating); Source A removed

src_c_00 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: For years, I tried to figure out why you were never around. I convinced myself you were trying your hardest. But now I realize the truth. You were just as lost, weren't you? I never knew you were so hollow inside.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.920

blend 7.33 · genu 1.92
invWER 0.81 (WER 0.19)

reward 0.855

blend 5.17 · genu 1.57
invWER 1.00 (WER 0.00)

reward 0.688

blend 3.16 · genu 1.93
invWER 0.95 (WER 0.05)

bottom

reward 0.263

blend 4.42 · genu 1.94
invWER 0.31 (WER 0.69)

reward 0.010

blend 1.66 · genu 1.20
invWER 0.02 (WER 0.98)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.890

blend 7.33 · genu 1.88
invWER 0.81 (WER 0.19)

reward 0.809

blend 5.17 · genu 1.44
invWER 1.00 (WER 0.00)

reward 0.583

blend 3.16 · genu 1.46
invWER 0.95 (WER 0.05)

bottom

reward 0.227

blend 4.42 · genu 1.44
invWER 0.31 (WER 0.69)

reward 0.008

blend 1.66 · genu 0.75
invWER 0.02 (WER 0.98)

src_b_00 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Oh, but how am I even supposed to say it? Even if I did, I mean, w-w-what would happen next, right? Like, w-w-what if you hated it a-a-and you kicked me o-

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.016

blend 7.88 · genu 3.19
invWER 0.71 (WER 0.29)

reward 0.933

blend 5.75 · genu 2.26
invWER 0.89 (WER 0.10)

reward 0.931

blend 6.56 · genu 2.35
invWER 0.82 (WER 0.18)

bottom

reward 0.098

blend 4.92 · genu 3.73
invWER 0.08 (WER 0.92)

reward 0.068

blend 4.96 · genu 1.74
invWER 0.08 (WER 0.92)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.914

blend 6.56 · genu 2.46
invWER 0.82 (WER 0.18)

reward 0.880

blend 7.88 · genu 2.38
invWER 0.71 (WER 0.29)

reward 0.873

blend 5.75 · genu 2.08
invWER 0.89 (WER 0.10)

bottom

reward 0.106

blend 4.92 · genu 4.77
invWER 0.08 (WER 0.92)

reward 0.067

blend 4.96 · genu 1.80
invWER 0.08 (WER 0.92)

german_src_00 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: herzlich willkommen bei den dsa nachrichten in 21 minuten, präsentiert von hintern im auge, eurer wöchentlichen nachrichtensendung rund um dsa, die schwarze katze und aventuria. heute geht es um ein update zur sonnenküste und die gerade erschienenen pdfs der zehnten welle des collectors club.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.508

blend 2.76 · genu 1.52
invWER 0.84 (WER 0.16)

reward 0.508

blend 1.12 · genu 2.07
invWER 0.93 (WER 0.07)

reward 0.421

blend 1.50 · genu 1.63
invWER 0.84 (WER 0.16)

bottom

reward 0.104

blend 0.43 · genu 0.14
invWER 0.93 (WER 0.07)

reward 0.067

blend 0.13 · genu 0.09
invWER 0.93 (WER 0.07)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.277

blend 2.76 · genu 0.03
invWER 0.84 (WER 0.16)

reward 0.195

blend 1.50 · genu 0.19
invWER 0.84 (WER 0.16)

reward 0.175

blend 1.12 · genu 0.16
invWER 0.93 (WER 0.07)

bottom

reward 0.071

blend 0.43 · genu -0.09
invWER 0.93 (WER 0.07)

reward 0.046

blend 0.13 · genu -0.07
invWER 0.93 (WER 0.07)

src_d_00 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: My apologies, sorry—alright, I just don't have that figure handy at the moment. Can we move past this? Yeah, let's move on. I always do this. I talk right over the silent part of my own mind. It isn't the show. It's me.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.452

blend 2.80 · genu 0.85
invWER 0.94 (WER 0.06)

reward 0.278

blend 6.86 · genu 1.19
invWER 0.29 (WER 0.71)

reward 0.192

blend 8.12 · genu 1.58
invWER 0.17 (WER 0.83)

bottom

reward 0.000

blend 4.36 · genu 1.64
invWER 0.00 (WER 1.02)

reward 0.000

blend 2.45 · genu 1.01
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.728

blend 2.80 · genu 2.65
invWER 0.94 (WER 0.06)

reward 0.332

blend 6.86 · genu 2.38
invWER 0.29 (WER 0.71)

reward 0.225

blend 8.12 · genu 2.88
invWER 0.17 (WER 0.83)

bottom

reward 0.000

blend 4.36 · genu 3.95
invWER 0.00 (WER 1.02)

reward 0.000

blend 2.45 · genu 1.43
invWER 0.00 (WER 1.00)

src_c_01 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Say this is a lie. Answer me... Why would you do this? We were going to... I don't deserve this! I won't marry you. Just leave. Please. I can't do this...

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.569

blend 9.00 · genu 1.85
invWER 0.44 (WER 0.56)

reward 0.286

blend 7.97 · genu 1.99
invWER 0.24 (WER 0.77)

reward 0.226

blend 5.92 · genu 1.74
invWER 0.24 (WER 0.77)

bottom

reward 0.018

blend 1.87 · genu 2.09
invWER 0.03 (WER 0.97)

reward 0.007

blend 0.72 · genu 0.65
invWER 0.03 (WER 0.97)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.531

blend 9.00 · genu 1.51
invWER 0.44 (WER 0.56)

reward 0.260

blend 7.97 · genu 1.53
invWER 0.24 (WER 0.77)

reward 0.224

blend 5.92 · genu 1.84
invWER 0.24 (WER 0.77)

bottom

reward 0.012

blend 1.87 · genu 1.11
invWER 0.03 (WER 0.97)

reward 0.008

blend 0.72 · genu 0.92
invWER 0.03 (WER 0.97)

src_b_01 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: I really appreciate everyone for the kind messages you've sent me lately. They seriously... Man, thank goodness my moderators do such an amazing job.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.970

blend 2.84 · genu 3.42
invWER 1.00 (WER 0.00)

reward 0.827

blend 1.98 · genu 3.12
invWER 1.00 (WER 0.00)

reward 0.780

blend 3.73 · genu 2.11
invWER 0.96 (WER 0.04)

bottom

reward 0.604

blend 2.79 · genu 3.64
invWER 0.60 (WER 0.40)

reward 0.530

blend 0.00 · genu 2.71
invWER 0.96 (WER 0.04)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.913

blend 2.84 · genu 3.43
invWER 1.00 (WER 0.00)

reward 0.845

blend 3.73 · genu 2.71
invWER 0.96 (WER 0.04)

reward 0.821

blend 3.73 · genu 2.57
invWER 0.96 (WER 0.04)

bottom

reward 0.467

blend 2.79 · genu 2.67
invWER 0.60 (WER 0.40)

reward 0.462

blend 0.00 · genu 2.56
invWER 0.96 (WER 0.04)

german_src_01 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: zudem gibt es nachschub für foundry wtt. darüber hinaus werden die ulysses classics ausschließlich im crowdfunding angeboten.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.719

blend 3.37 · genu 2.04
invWER 0.94 (WER 0.06)

reward 0.652

blend 2.90 · genu 2.16
invWER 0.88 (WER 0.12)

reward 0.588

blend 3.60 · genu 1.40
invWER 0.88 (WER 0.12)

bottom

reward 0.157

blend 0.31 · genu 0.49
invWER 0.94 (WER 0.06)

reward 0.063

blend 0.00 · genu 0.18
invWER 0.82 (WER 0.18)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.356

blend 3.37 · genu -0.05
invWER 0.94 (WER 0.06)

reward 0.351

blend 3.60 · genu -0.07
invWER 0.88 (WER 0.12)

reward 0.293

blend 2.90 · genu -0.04
invWER 0.88 (WER 0.12)

bottom

reward 0.065

blend 0.31 · genu -0.06
invWER 0.94 (WER 0.06)

reward 0.033

blend 0.00 · genu -0.05
invWER 0.82 (WER 0.18)

src_d_02 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Hey... are you up? I need to confess something I've been hiding. I'm terrified, but I'm done being afraid. I just hope you understand. I love you too much to keep living a lie. That's it.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.850

blend 5.83 · genu 1.99
invWER 0.85 (WER 0.15)

reward 0.798

blend 8.14 · genu 1.97
invWER 0.65 (WER 0.35)

reward 0.527

blend 2.21 · genu 1.63
invWER 0.93 (WER 0.07)

bottom

reward 0.245

blend 5.88 · genu 1.38
invWER 0.28 (WER 0.72)

reward 0.201

blend 9.51 · genu 1.85
invWER 0.15 (WER 0.85)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.688

blend 5.83 · genu 1.05
invWER 0.85 (WER 0.15)

reward 0.688

blend 8.14 · genu 1.16
invWER 0.65 (WER 0.35)

reward 0.440

blend 8.72 · genu 1.49
invWER 0.38 (WER 0.62)

bottom

reward 0.222

blend 5.88 · genu 1.00
invWER 0.28 (WER 0.72)

reward 0.190

blend 9.51 · genu 1.59
invWER 0.15 (WER 0.85)

src_c_02 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Say this is a lie. Tell me... Why would you do this? We had a future... I am worth more than this! I won't marry you. Leave. Please. Just go.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.195

blend 6.33 · genu 1.76
invWER 0.19 (WER 0.81)

reward 0.183

blend 7.53 · genu 1.81
invWER 0.16 (WER 0.84)

reward 0.152

blend 3.78 · genu 1.94
invWER 0.19 (WER 0.81)

bottom

reward 0.012

blend 0.93 · genu 1.17
invWER 0.03 (WER 0.97)

reward 0.000

blend 2.28 · genu 1.45
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.188

blend 7.53 · genu 2.16
invWER 0.16 (WER 0.84)

reward 0.175

blend 6.33 · genu 1.33
invWER 0.19 (WER 0.81)

reward 0.127

blend 3.78 · genu 1.36
invWER 0.19 (WER 0.81)

bottom

reward 0.011

blend 1.64 · genu 0.80
invWER 0.03 (WER 0.97)

reward 0.000

blend 2.28 · genu 1.24
invWER 0.00 (WER 1.00)

src_b_02 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Large, frightened eyes. Well, large from your perspective, at least.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.024

blend 5.66 · genu 2.21
invWER 1.00 (WER 0.00)

reward 1.006

blend 5.77 · genu 2.05
invWER 1.00 (WER 0.00)

reward 0.941

blend 4.80 · genu 2.22
invWER 1.00 (WER 0.00)

bottom

reward 0.733

blend 2.46 · genu 2.80
invWER 0.90 (WER 0.10)

reward 0.699

blend 1.81 · genu 2.94
invWER 0.90 (WER 0.10)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.908

blend 4.18 · genu 2.61
invWER 1.00 (WER 0.00)

reward 0.845

blend 5.66 · genu 1.36
invWER 1.00 (WER 0.00)

reward 0.835

blend 5.77 · genu 1.24
invWER 1.00 (WER 0.00)

bottom

reward 0.524

blend 2.46 · genu 1.70
invWER 0.90 (WER 0.10)

reward 0.503

blend 1.81 · genu 1.94
invWER 0.90 (WER 0.10)

german_src_02 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: schlangengrab, die fünfte folge der hörspielreihe phileasson, wurde veröffentlicht.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.888

blend 6.12 · genu 2.59
invWER 0.78 (WER 0.22)

reward 0.511

blend 2.90 · genu 1.72
invWER 0.78 (WER 0.22)

reward 0.501

blend 2.38 · genu 1.17
invWER 1.00 (WER 0.00)

bottom

reward 0.098

blend 0.18 · genu 0.35
invWER 0.78 (WER 0.22)

reward 0.093

blend 0.20 · genu 0.30
invWER 0.78 (WER 0.22)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.678

blend 6.12 · genu 1.25
invWER 0.78 (WER 0.22)

reward 0.341

blend 1.90 · genu 1.19
invWER 0.78 (WER 0.22)

reward 0.329

blend 2.90 · genu 0.50
invWER 0.78 (WER 0.22)

bottom

reward 0.080

blend 0.20 · genu 0.20
invWER 0.78 (WER 0.22)

reward 0.046

blend 0.18 · genu -0.05
invWER 0.78 (WER 0.22)

src_d_03 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Hi, my love. No, no, I'm fine, really. It was just a long day, that's all. But sometimes... sometimes the silence feels so heavy. I don't know how to fill it.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.936

blend 6.91 · genu 1.73
invWER 0.88 (WER 0.12)

reward 0.867

blend 6.61 · genu 1.48
invWER 0.88 (WER 0.12)

reward 0.833

blend 6.79 · genu 1.74
invWER 0.79 (WER 0.21)

bottom

reward 0.000

blend 0.15 · genu 1.20
invWER 0.00 (WER 1.00)

reward 0.000

blend 6.98 · genu 1.82
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.854

blend 6.91 · genu 1.35
invWER 0.88 (WER 0.12)

reward 0.851

blend 6.61 · genu 1.50
invWER 0.88 (WER 0.12)

reward 0.841

blend 6.79 · genu 1.96
invWER 0.79 (WER 0.21)

bottom

reward 0.000

blend 0.15 · genu 2.31
invWER 0.00 (WER 1.00)

reward 0.000

blend 6.98 · genu 1.56
invWER 0.00 (WER 1.00)

src_c_03 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Say this is a lie. Say it... How could you do this? We had plans... I am worth more than this! I'm not marrying you. Leave. Please. I just can't.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.324

blend 6.66 · genu 1.24
invWER 0.34 (WER 0.66)

reward 0.215

blend 9.84 · genu 1.87
invWER 0.16 (WER 0.84)

reward 0.213

blend 7.59 · genu 1.79
invWER 0.19 (WER 0.81)

bottom

reward 0.057

blend 2.31 · genu 1.76
invWER 0.09 (WER 0.91)

reward 0.027

blend 5.27 · genu 1.60
invWER 0.03 (WER 0.97)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.338

blend 6.66 · genu 1.59
invWER 0.34 (WER 0.66)

reward 0.215

blend 9.84 · genu 2.03
invWER 0.16 (WER 0.84)

reward 0.203

blend 7.59 · genu 1.63
invWER 0.19 (WER 0.81)

bottom

reward 0.053

blend 2.31 · genu 1.70
invWER 0.09 (WER 0.91)

reward 0.026

blend 5.27 · genu 1.60
invWER 0.03 (WER 0.97)

src_b_03 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: <sighs> I have to admit, though, you look adorable and incredibly happy.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.977

blend 6.24 · genu 2.12
invWER 0.92 (WER 0.08)

reward 0.850

blend 4.19 · genu 2.47
invWER 0.92 (WER 0.08)

reward 0.754

blend 5.28 · genu 1.78
invWER 0.83 (WER 0.17)

bottom

reward 0.340

blend 5.13 · genu 1.38
invWER 0.42 (WER 0.58)

reward 0.236

blend 1.23 · genu 1.00
invWER 0.67 (WER 0.33)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.855

blend 6.24 · genu 1.54
invWER 0.92 (WER 0.08)

reward 0.751

blend 5.28 · genu 1.92
invWER 0.83 (WER 0.17)

reward 0.664

blend 5.12 · genu 1.40
invWER 0.83 (WER 0.17)

bottom

reward 0.393

blend 5.13 · genu 2.26
invWER 0.42 (WER 0.58)

reward 0.198

blend 1.23 · genu 0.75
invWER 0.67 (WER 0.33)

german_src_03 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: außerdem gibt es etwas neues im skripturium. weitere infos zu den produkten findest du in der passenden quellenangabe oder über die links in der beschreibung zum video oder podcast. und nun viel spaß mit den nachrichten. fangen wir zuerst mit den neuerscheinungen an.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.896

blend 5.55 · genu 2.51
invWER 0.84 (WER 0.16)

reward 0.760

blend 3.55 · genu 2.23
invWER 0.93 (WER 0.07)

reward 0.710

blend 2.91 · genu 2.09
invWER 0.98 (WER 0.02)

bottom

reward 0.102

blend 0.60 · genu 0.02
invWER 0.95 (WER 0.05)

reward 0.089

blend 0.45 · genu 0.02
invWER 0.98 (WER 0.02)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.804

blend 5.55 · genu 2.11
invWER 0.84 (WER 0.16)

reward 0.532

blend 3.55 · genu 1.00
invWER 0.93 (WER 0.07)

reward 0.377

blend 2.44 · genu 0.55
invWER 0.98 (WER 0.02)

bottom

reward 0.098

blend 0.60 · genu -0.04
invWER 0.95 (WER 0.05)

reward 0.091

blend 0.45 · genu -0.01
invWER 0.98 (WER 0.02)

src_d_04 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Wonderful speech, darling. Very elegant. You certainly know how to charm a crowd. I recall winning that exact same prize... three consecutive years. But things move on. You just sit there smiling and applauding, don't you?

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.764

blend 5.96 · genu 1.16
invWER 0.89 (WER 0.11)

reward 0.457

blend 7.02 · genu 1.67
invWER 0.43 (WER 0.57)

reward 0.403

blend 8.05 · genu 1.59
invWER 0.35 (WER 0.65)

bottom

reward 0.125

blend 6.16 · genu 1.42
invWER 0.14 (WER 0.86)

reward 0.000

blend 1.36 · genu 0.67
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.732

blend 5.96 · genu 1.04
invWER 0.89 (WER 0.11)

reward 0.391

blend 7.02 · genu 0.90
invWER 0.43 (WER 0.57)

reward 0.339

blend 8.05 · genu 0.65
invWER 0.35 (WER 0.65)

bottom

reward 0.098

blend 3.80 · genu 0.52
invWER 0.19 (WER 0.81)

reward 0.000

blend 1.36 · genu 1.02
invWER 0.00 (WER 1.00)

src_c_04 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: You took everything. Every moment I wasted on you was me fooling myself. I wished you were dead. But I can't... I can't even picture who I'd be without resenting you. Why do you still matter?

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.470

blend 1.35 · genu 2.53
invWER 0.72 (WER 0.28)

reward 0.433

blend 3.74 · genu 2.06
invWER 0.54 (WER 0.46)

reward 0.399

blend 4.57 · genu 1.94
invWER 0.46 (WER 0.54)

bottom

reward 0.233

blend 0.57 · genu 2.69
invWER 0.38 (WER 0.61)

reward 0.212

blend 1.49 · genu 1.57
invWER 0.44 (WER 0.56)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.358

blend 3.74 · genu 1.43
invWER 0.54 (WER 0.46)

reward 0.342

blend 4.57 · genu 1.40
invWER 0.46 (WER 0.54)

reward 0.339

blend 1.35 · genu 1.71
invWER 0.72 (WER 0.28)

bottom

reward 0.158

blend 1.49 · genu 0.98
invWER 0.44 (WER 0.56)

reward 0.103

blend 0.57 · genu 0.96
invWER 0.38 (WER 0.61)

src_b_04 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: You look incredibly gorgeous and feminine. <giggles> Yeah, I'm really happy you're more comfortable this way.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.887

blend 2.79 · genu 3.28
invWER 0.94 (WER 0.06)

reward 0.883

blend 3.36 · genu 2.96
invWER 0.94 (WER 0.06)

reward 0.862

blend 4.07 · genu 2.46
invWER 0.94 (WER 0.06)

bottom

reward 0.465

blend 4.88 · genu 2.74
invWER 0.44 (WER 0.56)

reward 0.245

blend 4.24 · genu 1.43
invWER 0.33 (WER 0.67)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.658

blend 4.07 · genu 1.43
invWER 0.94 (WER 0.06)

reward 0.589

blend 4.16 · genu 1.17
invWER 0.89 (WER 0.11)

reward 0.572

blend 3.36 · genu 1.31
invWER 0.94 (WER 0.06)

bottom

reward 0.342

blend 4.88 · genu 1.38
invWER 0.44 (WER 0.56)

reward 0.232

blend 4.24 · genu 1.32
invWER 0.33 (WER 0.67)

german_src_04 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: wie eben schon erwähnt, ist schlangenrabe, der fünfte teil der filhassonsage, jetzt als hörbuch erschienen.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.417

blend 8.56 · genu 3.91
invWER 0.87 (WER 0.13)

reward 1.072

blend 6.58 · genu 2.39
invWER 0.93 (WER 0.07)

reward 0.676

blend 4.11 · genu 2.08
invWER 0.80 (WER 0.20)

bottom

reward 0.458

blend 3.18 · genu 1.40
invWER 0.73 (WER 0.27)

reward 0.325

blend 1.75 · genu 0.84
invWER 0.87 (WER 0.13)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 1.234

blend 8.56 · genu 3.07
invWER 0.87 (WER 0.13)

reward 0.774

blend 6.58 · genu 0.72
invWER 0.93 (WER 0.07)

reward 0.393

blend 3.79 · genu 0.15
invWER 0.87 (WER 0.13)

bottom

reward 0.327

blend 3.65 · genu -0.03
invWER 0.80 (WER 0.20)

reward 0.267

blend 1.75 · genu 0.50
invWER 0.87 (WER 0.13)

src_d_05 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: I never… I've never actually seen that before. It's like a miniature world, simply resting there. How long had it been around? And now it has vanished. You stole it. Exactly like all the rest. Every last piece.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.163

blend 8.28 · genu 1.55
invWER 1.00 (WER 0.00)

reward 0.935

blend 5.72 · genu 1.71
invWER 1.00 (WER 0.00)

reward 0.917

blend 8.25 · genu 1.30
invWER 0.82 (WER 0.17)

bottom

reward 0.387

blend 5.99 · genu 1.16
invWER 0.45 (WER 0.55)

reward 0.000

blend 2.65 · genu 0.74
invWER 0.00 (WER 1.07)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 1.051

blend 8.28 · genu 1.03
invWER 1.00 (WER 0.00)

reward 0.940

blend 7.58 · genu 1.24
invWER 0.93 (WER 0.07)

reward 0.896

blend 8.25 · genu 1.26
invWER 0.82 (WER 0.17)

bottom

reward 0.362

blend 5.99 · genu 0.93
invWER 0.45 (WER 0.55)

reward 0.000

blend 2.65 · genu 0.71
invWER 0.00 (WER 1.07)

src_c_05 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: I made the sandwich you love. Plenty of pickles, exactly the way you prefer. Make sure to write, alright? Or call whenever you're able. I understand you must leave. I just... really need you to return safely. Give me your word. You are so courageous, sweetheart. I will be waiting right here.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.259

blend 3.52 · genu 1.75
invWER 0.36 (WER 0.64)

reward 0.255

blend 6.92 · genu 1.63
invWER 0.25 (WER 0.76)

reward 0.249

blend 5.32 · genu 1.95
invWER 0.26 (WER 0.74)

bottom

reward 0.004

blend 0.95 · genu 0.49
invWER 0.02 (WER 0.98)

reward 0.000

blend 4.77 · genu 1.31
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.254

blend 6.92 · genu 1.75
invWER 0.25 (WER 0.76)

reward 0.219

blend 3.52 · genu 1.24
invWER 0.36 (WER 0.64)

reward 0.199

blend 5.32 · genu 1.01
invWER 0.26 (WER 0.74)

bottom

reward 0.006

blend 1.16 · genu 0.82
invWER 0.02 (WER 0.98)

reward 0.000

blend 4.77 · genu 1.68
invWER 0.00 (WER 1.00)

src_b_05 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Look, you're already leaning your head into my hands. You really can't hide how much you enjoy getting scratches, can you?

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.642

blend 2.30 · genu 2.29
invWER 0.91 (WER 0.09)

reward 0.637

blend 3.00 · genu 2.51
invWER 0.78 (WER 0.22)

reward 0.536

blend 1.51 · genu 1.82
invWER 1.00 (WER 0.00)

bottom

reward 0.365

blend 1.27 · genu 1.04
invWER 1.00 (WER 0.00)

reward 0.359

blend 1.31 · genu 0.98
invWER 1.00 (WER 0.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.432

blend 2.30 · genu 1.15
invWER 0.91 (WER 0.09)

reward 0.395

blend 3.00 · genu 0.92
invWER 0.78 (WER 0.22)

reward 0.333

blend 2.36 · genu 0.28
invWER 1.00 (WER 0.00)

bottom

reward 0.217

blend 1.34 · genu 0.21
invWER 1.00 (WER 0.00)

reward 0.142

blend 0.31 · genu 0.37
invWER 1.00 (WER 0.00)

german_src_05 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: für foundry vtt gibt es verschiedene neuigkeiten. zum einen das aventurische compendium 1. dieses erschien am 16.07. zum preis von 17,95. wie immer nur im f shop erhältlich. außerdem gehört das szenario midsummer rache dazu. achtung, spoiler!

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.950

blend 3.92 · genu 3.49
invWER 0.87 (WER 0.13)

reward 0.865

blend 4.80 · genu 2.50
invWER 0.87 (WER 0.13)

reward 0.810

blend 4.41 · genu 2.09
invWER 0.92 (WER 0.08)

bottom

reward 0.532

blend 3.18 · genu 1.53
invWER 0.82 (WER 0.18)

reward 0.469

blend 2.37 · genu 1.76
invWER 0.77 (WER 0.23)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.640

blend 4.41 · genu 1.21
invWER 0.92 (WER 0.08)

reward 0.613

blend 3.92 · genu 1.55
invWER 0.87 (WER 0.13)

reward 0.492

blend 4.80 · genu 0.21
invWER 0.87 (WER 0.13)

bottom

reward 0.284

blend 2.75 · genu 0.01
invWER 0.87 (WER 0.13)

reward 0.252

blend 2.38 · genu 0.12
invWER 0.82 (WER 0.18)

src_d_06 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: So you really consider that an obscure track? That's adorable. It's as basic as ordering plain water at a jazz club. I can't believe you brought that in here. No way. And then she spun that exact same tune... and I was left completely speechless. I guess some albums stay permanently damaged. It's amusing how the stylus always drops right back into the same old track.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.379

blend 4.92 · genu 1.07
invWER 0.51 (WER 0.49)

reward 0.368

blend 3.52 · genu 1.24
invWER 0.59 (WER 0.41)

reward 0.148

blend 6.13 · genu 1.11
invWER 0.17 (WER 0.83)

bottom

reward 0.050

blend 5.72 · genu 1.34
invWER 0.06 (WER 0.94)

reward 0.021

blend 1.85 · genu 1.39
invWER 0.04 (WER 0.96)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.485

blend 4.92 · genu 2.38
invWER 0.51 (WER 0.49)

reward 0.439

blend 3.52 · genu 2.06
invWER 0.59 (WER 0.41)

reward 0.164

blend 6.13 · genu 1.76
invWER 0.17 (WER 0.83)

bottom

reward 0.050

blend 5.72 · genu 1.50
invWER 0.06 (WER 0.94)

reward 0.024

blend 1.85 · genu 1.97
invWER 0.04 (WER 0.96)

src_c_06 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: The hallway lights glow like tiny moons. I feel like I can finally find peace. Oh, damn it. Why did it take so much time? I've loved you from the start. Forever.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.247

blend 5.17 · genu 1.84
invWER 0.27 (WER 0.73)

reward 0.186

blend 1.38 · genu 1.21
invWER 0.45 (WER 0.55)

reward 0.174

blend 2.01 · genu 1.25
invWER 0.36 (WER 0.64)

bottom

reward 0.009

blend 0.11 · genu 1.38
invWER 0.03 (WER 0.97)

reward 0.007

blend 0.48 · genu 0.70
invWER 0.03 (WER 0.97)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.239

blend 5.17 · genu 1.84
invWER 0.27 (WER 0.73)

reward 0.160

blend 1.38 · genu 0.97
invWER 0.45 (WER 0.55)

reward 0.153

blend 2.01 · genu 1.01
invWER 0.36 (WER 0.64)

bottom

reward 0.008

blend 0.48 · genu 1.04
invWER 0.03 (WER 0.97)

reward 0.008

blend 0.11 · genu 1.16
invWER 0.03 (WER 0.97)

src_b_06 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Yours, though? Oh, easily the greatest. But Ria...

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.969

blend 6.61 · genu 2.15
invWER 0.88 (WER 0.12)

reward 0.960

blend 6.30 · genu 2.26
invWER 0.88 (WER 0.12)

reward 0.827

blend 6.45 · genu 2.20
invWER 0.75 (WER 0.25)

bottom

reward 0.307

blend 3.22 · genu 2.42
invWER 0.38 (WER 0.62)

reward 0.000

blend 0.00 · genu 2.54
invWER 0.00 (WER 6.75)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.853

blend 6.61 · genu 1.57
invWER 0.88 (WER 0.12)

reward 0.823

blend 6.30 · genu 1.55
invWER 0.88 (WER 0.12)

reward 0.785

blend 6.45 · genu 2.08
invWER 0.75 (WER 0.25)

bottom

reward 0.270

blend 3.22 · genu 2.07
invWER 0.38 (WER 0.62)

reward 0.000

blend 0.00 · genu 1.00
invWER 0.00 (WER 6.75)

german_src_06 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: ich betrachte hier jetzt nämlich ebenfalls einige screenshots aus dem abenteuer selbst.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.004

blend 4.58 · genu 3.16
invWER 0.92 (WER 0.08)

reward 0.900

blend 4.04 · genu 2.41
invWER 1.00 (WER 0.00)

reward 0.781

blend 3.33 · genu 2.16
invWER 1.00 (WER 0.00)

bottom

reward 0.460

blend 1.34 · genu 1.50
invWER 1.00 (WER 0.00)

reward 0.342

blend 0.84 · genu 1.31
invWER 0.92 (WER 0.08)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.877

blend 4.58 · genu 2.67
invWER 0.92 (WER 0.08)

reward 0.816

blend 4.04 · genu 2.15
invWER 1.00 (WER 0.00)

reward 0.662

blend 3.04 · genu 1.83
invWER 1.00 (WER 0.00)

bottom

reward 0.441

blend 1.34 · genu 1.53
invWER 1.00 (WER 0.00)

reward 0.420

blend 1.79 · genu 1.13
invWER 1.00 (WER 0.00)

src_d_07 Source D 🇺🇸 EN <5 variants

Reference (original speaker)

Paraphrased text spoken: I paid in blood for all these scars. Yet you think you've won. But you left your side exposed, idiot. And my aim never fails.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.929

blend 6.56 · genu 1.64
invWER 0.92 (WER 0.08)

reward 0.714

blend 5.10 · genu 1.01
invWER 0.96 (WER 0.04)

reward 0.340

blend 1.69 · genu 1.11
invWER 0.81 (WER 0.19)

bottom

reward 0.113

blend 4.21 · genu 1.44
invWER 0.15 (WER 0.85)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.819

blend 6.56 · genu 1.08
invWER 0.92 (WER 0.08)

reward 0.693

blend 5.10 · genu 0.96
invWER 0.96 (WER 0.04)

reward 0.302

blend 1.69 · genu 0.92
invWER 0.81 (WER 0.19)

bottom

reward 0.105

blend 4.21 · genu 1.25
invWER 0.15 (WER 0.85)

src_c_07 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: You've displayed extraordinary... resilience. Over these three years... I haven't seen a struggle quite like yours. I hoped to rescue you. I just... I had to make sure you knew you were important.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.745

blend 7.18 · genu 1.98
invWER 0.66 (WER 0.34)

reward 0.536

blend 4.53 · genu 1.70
invWER 0.66 (WER 0.34)

reward 0.484

blend 5.98 · genu 1.89
invWER 0.49 (WER 0.51)

bottom

reward 0.019

blend 3.45 · genu 1.52
invWER 0.03 (WER 0.97)

reward 0.000

blend 3.07 · genu 1.54
invWER 0.00 (WER 1.29)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.714

blend 7.18 · genu 1.89
invWER 0.66 (WER 0.34)

reward 0.437

blend 4.53 · genu 0.97
invWER 0.66 (WER 0.34)

reward 0.433

blend 5.98 · genu 1.45
invWER 0.49 (WER 0.51)

bottom

reward 0.017

blend 1.60 · genu 0.51
invWER 0.06 (WER 0.94)

reward 0.000

blend 3.07 · genu 0.98
invWER 0.00 (WER 1.29)

src_b_07 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: What can I do to make you give in and let me pet you? Come on, there has to be something you'd take in return.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.042

blend 5.89 · genu 2.18
invWER 1.00 (WER 0.00)

reward 0.887

blend 5.62 · genu 1.69
invWER 0.96 (WER 0.04)

reward 0.803

blend 5.45 · genu 1.32
invWER 0.96 (WER 0.04)

bottom

reward 0.280

blend 0.63 · genu 0.92
invWER 1.00 (WER 0.00)

reward 0.254

blend 0.74 · genu 0.73
invWER 1.00 (WER 0.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.879

blend 5.89 · genu 1.43
invWER 1.00 (WER 0.00)

reward 0.716

blend 5.45 · genu 0.89
invWER 0.96 (WER 0.04)

reward 0.714

blend 5.62 · genu 0.78
invWER 0.96 (WER 0.04)

bottom

reward 0.253

blend 0.74 · genu 0.77
invWER 1.00 (WER 0.00)

reward 0.227

blend 0.63 · genu 0.68
invWER 1.00 (WER 0.00)

german_src_07 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: das hier ist das abenteuer zur torwahl im heldenwerkformat. es erschien außerdem am 16.07. für 6,99 und ihr könnt es regulär im fshop erwerben.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.255

blend 6.63 · genu 3.48
invWER 0.92 (WER 0.08)

reward 1.107

blend 5.70 · genu 4.03
invWER 0.81 (WER 0.19)

reward 0.916

blend 5.84 · genu 2.70
invWER 0.81 (WER 0.19)

bottom

reward 0.434

blend 2.41 · genu 1.35
invWER 0.81 (WER 0.19)

reward 0.212

blend 0.43 · genu 0.81
invWER 0.88 (WER 0.12)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.784

blend 6.63 · genu 0.82
invWER 0.92 (WER 0.08)

reward 0.741

blend 5.70 · genu 1.77
invWER 0.81 (WER 0.19)

reward 0.624

blend 5.84 · genu 0.83
invWER 0.81 (WER 0.19)

bottom

reward 0.216

blend 2.13 · genu -0.04
invWER 0.85 (WER 0.15)

reward 0.070

blend 0.43 · genu -0.07
invWER 0.88 (WER 0.12)

src_d_08 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Well, what do you know... check this out. Hidden behind that rusty gearbox all these months. Aha! There you are, you gorgeous piece of scrap! I promised I'd track it down, kid. I told you so!

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.589

blend 7.32 · genu 1.68
invWER 0.54 (WER 0.46)

reward 0.183

blend 1.66 · genu 1.46
invWER 0.38 (WER 0.62)

reward 0.139

blend 5.02 · genu 1.66
invWER 0.16 (WER 0.84)

bottom

reward 0.023

blend 2.04 · genu 0.99
invWER 0.05 (WER 0.95)

reward 0.000

blend 1.89 · genu 2.14
invWER 0.00 (WER 1.14)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.578

blend 7.32 · genu 1.71
invWER 0.54 (WER 0.46)

reward 0.169

blend 1.66 · genu 1.37
invWER 0.38 (WER 0.62)

reward 0.137

blend 0.04 · genu 0.99
invWER 0.62 (WER 0.38)

bottom

reward 0.022

blend 2.04 · genu 0.92
invWER 0.05 (WER 0.95)

reward 0.000

blend 1.89 · genu 1.46
invWER 0.00 (WER 1.14)

src_c_08 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: I've kept this for... wow, nearly a decade now. I kept waiting for the right time to come along. I always wanted it to be yours. I just didn't realize parting with it would hurt this badly. You deserve every bit of happiness out there.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.755

blend 4.16 · genu 1.95
invWER 0.91 (WER 0.09)

reward 0.726

blend 3.31 · genu 2.23
invWER 0.91 (WER 0.09)

reward 0.454

blend 1.75 · genu 1.42
invWER 0.94 (WER 0.06)

bottom

reward 0.048

blend 1.61 · genu 1.91
invWER 0.09 (WER 0.92)

reward 0.042

blend 3.52 · genu 1.39
invWER 0.06 (WER 0.94)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.687

blend 4.16 · genu 1.69
invWER 0.91 (WER 0.09)

reward 0.669

blend 3.31 · genu 2.08
invWER 0.91 (WER 0.09)

reward 0.518

blend 1.75 · genu 1.95
invWER 0.94 (WER 0.06)

bottom

reward 0.043

blend 1.61 · genu 1.76
invWER 0.09 (WER 0.92)

reward 0.041

blend 3.52 · genu 1.46
invWER 0.06 (WER 0.94)

src_b_08 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Why? <laughs> What makes me love you? Oh, honey, i... I don't really need a reason to love you.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.836

blend 6.33 · genu 1.64
invWER 0.85 (WER 0.15)

reward 0.761

blend 5.76 · genu 1.47
invWER 0.85 (WER 0.15)

reward 0.737

blend 5.17 · genu 1.64
invWER 0.85 (WER 0.15)

bottom

reward 0.276

blend 3.56 · genu 1.14
invWER 0.45 (WER 0.55)

reward 0.000

blend 5.76 · genu 2.57
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.803

blend 6.33 · genu 1.55
invWER 0.85 (WER 0.15)

reward 0.761

blend 5.76 · genu 1.60
invWER 0.85 (WER 0.15)

reward 0.694

blend 5.17 · genu 1.49
invWER 0.85 (WER 0.15)

bottom

reward 0.323

blend 3.56 · genu 1.85
invWER 0.45 (WER 0.55)

reward 0.000

blend 5.76 · genu 0.79
invWER 0.00 (WER 1.00)

german_src_08 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: außerdem gibt es auch für unsere englischen freunde nachschub, und zwar mit dem aventuria bestiary, sprich dem ersten aventurischen bestiarium.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.511

blend 8.40 · genu 3.76
invWER 0.95 (WER 0.05)

reward 1.024

blend 4.85 · genu 2.92
invWER 0.95 (WER 0.05)

reward 0.912

blend 5.04 · genu 2.48
invWER 0.90 (WER 0.10)

bottom

reward 0.448

blend 3.58 · genu 0.84
invWER 0.80 (WER 0.20)

reward 0.434

blend 2.78 · genu 0.86
invWER 0.90 (WER 0.10)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 1.057

blend 8.40 · genu 1.32
invWER 0.95 (WER 0.05)

reward 0.708

blend 4.85 · genu 1.25
invWER 0.95 (WER 0.05)

reward 0.690

blend 5.04 · genu 1.26
invWER 0.90 (WER 0.10)

bottom

reward 0.389

blend 2.78 · genu 0.62
invWER 0.90 (WER 0.10)

reward 0.343

blend 3.58 · genu 0.13
invWER 0.80 (WER 0.20)

src_d_09 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Everything went perfectly. Completely effortless. Such a wonderful day. Oh. You didn't forget. I appreciate it.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.082

blend 8.90 · genu 2.45
invWER 0.06 (WER 0.94)

reward 0.039

blend 4.14 · genu 1.09
invWER 0.06 (WER 0.94)

reward 0.000

blend 4.19 · genu 1.10
invWER 0.00 (WER 1.00)

bottom

reward 0.000

blend 4.48 · genu 0.67
invWER 0.00 (WER 1.00)

reward 0.000

blend 5.04 · genu 2.89
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.074

blend 8.90 · genu 1.86
invWER 0.06 (WER 0.94)

reward 0.035

blend 4.14 · genu 0.81
invWER 0.06 (WER 0.94)

reward 0.000

blend 4.19 · genu 1.36
invWER 0.00 (WER 1.00)

bottom

reward 0.000

blend 4.48 · genu 0.96
invWER 0.00 (WER 1.00)

reward 0.000

blend 5.04 · genu 2.76
invWER 0.00 (WER 1.00)

src_c_09 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: I've been saving this for... gosh, nearly a decade now. I kept telling myself the right time would come. I always intended to give it to you. I just never realized how painful giving it up would be. You deserve every happiness in the world.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.449

blend 1.54 · genu 1.51
invWER 0.93 (WER 0.07)

reward 0.392

blend 3.21 · genu 2.24
invWER 0.50 (WER 0.50)

reward 0.366

blend 5.57 · genu 1.78
invWER 0.39 (WER 0.61)

bottom

reward 0.101

blend 0.44 · genu 1.44
invWER 0.28 (WER 0.72)

reward 0.016

blend 4.21 · genu 1.34
invWER 0.02 (WER 0.98)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.512

blend 1.54 · genu 2.04
invWER 0.93 (WER 0.07)

reward 0.346

blend 5.57 · genu 1.64
invWER 0.39 (WER 0.61)

reward 0.320

blend 3.21 · genu 1.60
invWER 0.50 (WER 0.50)

bottom

reward 0.112

blend 0.44 · genu 1.80
invWER 0.28 (WER 0.72)

reward 0.016

blend 4.21 · genu 1.53
invWER 0.02 (WER 0.98)

src_b_09 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: And just think if I, if I could only offer you my... Oh, man, Mimi, what are your thoughts?

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.106

blend 6.07 · genu 2.43
invWER 1.00 (WER 0.00)

reward 1.086

blend 6.00 · genu 3.44
invWER 0.84 (WER 0.16)

reward 1.064

blend 7.07 · genu 1.67
invWER 1.00 (WER 0.00)

bottom

reward 0.641

blend 5.61 · genu 2.68
invWER 0.58 (WER 0.42)

reward 0.431

blend 2.14 · genu 0.93
invWER 1.00 (WER 0.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.999

blend 7.07 · genu 1.44
invWER 1.00 (WER 0.00)

reward 0.978

blend 6.07 · genu 1.91
invWER 1.00 (WER 0.00)

reward 0.948

blend 6.00 · genu 2.82
invWER 0.84 (WER 0.16)

bottom

reward 0.462

blend 5.61 · genu 1.11
invWER 0.58 (WER 0.42)

reward 0.461

blend 2.14 · genu 1.17
invWER 1.00 (WER 0.00)

german_src_09 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: genauso wie das aventuria armory, also die aventurische rüstkammer, die ebenfalls für 19,99 dollar im internationalen f shop erschien.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.289

blend 8.01 · genu 3.13
invWER 0.90 (WER 0.10)

reward 1.113

blend 6.48 · genu 2.91
invWER 0.90 (WER 0.10)

reward 1.086

blend 6.82 · genu 2.57
invWER 0.90 (WER 0.10)

bottom

reward 0.757

blend 4.33 · genu 2.21
invWER 0.85 (WER 0.15)

reward 0.688

blend 4.47 · genu 1.46
invWER 0.90 (WER 0.10)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 1.102

blend 8.01 · genu 2.22
invWER 0.90 (WER 0.10)

reward 0.844

blend 6.48 · genu 1.43
invWER 0.90 (WER 0.10)

reward 0.822

blend 6.82 · genu 1.08
invWER 0.90 (WER 0.10)

bottom

reward 0.561

blend 4.46 · genu 0.76
invWER 0.90 (WER 0.10)

reward 0.531

blend 4.33 · genu 0.84
invWER 0.85 (WER 0.15)

src_d_11 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: You really think you're smart, don't you? That little trick we played— I'm still aching from it. No no no, I have to find the— stairs— everything went black—

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.804

blend 6.32 · genu 1.67
invWER 0.81 (WER 0.19)

reward 0.716

blend 6.63 · genu 1.12
invWER 0.78 (WER 0.22)

reward 0.517

blend 6.84 · genu 1.63
invWER 0.50 (WER 0.50)

bottom

reward 0.063

blend 6.03 · genu 1.95
invWER 0.06 (WER 0.94)

reward 0.000

blend 6.76 · genu 1.90
invWER 0.00 (WER 1.22)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.766

blend 6.32 · genu 1.55
invWER 0.81 (WER 0.19)

reward 0.762

blend 6.63 · genu 1.57
invWER 0.78 (WER 0.22)

reward 0.523

blend 6.84 · genu 1.85
invWER 0.50 (WER 0.50)

bottom

reward 0.061

blend 6.03 · genu 1.97
invWER 0.06 (WER 0.94)

reward 0.000

blend 6.76 · genu 1.19
invWER 0.00 (WER 1.22)

src_c_10 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: I promise I'll hunt them down. I'll make them suffer exactly what they put you through. I can't... I can't change what happened. I'm so sorry, sweetheart. I'm so sorry.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.618

blend 5.34 · genu 1.86
invWER 0.67 (WER 0.33)

reward 0.482

blend 3.04 · genu 2.01
invWER 0.67 (WER 0.33)

reward 0.283

blend 4.97 · genu 2.06
invWER 0.31 (WER 0.69)

bottom

reward 0.042

blend 2.06 · genu 1.34
invWER 0.08 (WER 0.92)

reward 0.034

blend 1.14 · genu 1.36
invWER 0.08 (WER 0.92)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.555

blend 5.34 · genu 1.47
invWER 0.67 (WER 0.33)

reward 0.360

blend 3.04 · genu 1.11
invWER 0.67 (WER 0.33)

reward 0.269

blend 4.97 · genu 1.98
invWER 0.31 (WER 0.69)

bottom

reward 0.035

blend 2.06 · genu 1.01
invWER 0.08 (WER 0.92)

reward 0.030

blend 1.14 · genu 1.18
invWER 0.08 (WER 0.92)

src_b_10 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: ...their presence hanging over you. And if you despise me, then I-

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.218

blend 6.32 · genu 2.89
invWER 1.00 (WER 0.00)

reward 0.857

blend 5.22 · genu 1.97
invWER 0.92 (WER 0.08)

reward 0.856

blend 3.25 · genu 2.60
invWER 1.00 (WER 0.00)

bottom

reward 0.626

blend 3.51 · genu 1.23
invWER 1.00 (WER 0.00)

reward 0.267

blend 2.38 · genu 0.63
invWER 0.67 (WER 0.33)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.944

blend 6.32 · genu 1.56
invWER 1.00 (WER 0.00)

reward 0.781

blend 7.58 · genu 1.39
invWER 0.75 (WER 0.25)

reward 0.775

blend 5.22 · genu 1.63
invWER 0.92 (WER 0.08)

bottom

reward 0.491

blend 3.25 · genu 0.70
invWER 1.00 (WER 0.00)

reward 0.258

blend 2.38 · genu 0.59
invWER 0.67 (WER 0.33)

german_src_10 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: außerdem gibt es neuigkeiten zu den hörspielen von winterzeit. folge 1, im kerker von gareth, ist seit dem 11. juli nämlich auch bei youtube verfügbar. natürlich völlig kostenfrei, mit einer laufzeit von 54 minuten 28.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.523

blend 8.52 · genu 4.38
invWER 0.89 (WER 0.11)

reward 1.234

blend 6.01 · genu 3.14
invWER 1.00 (WER 0.00)

reward 0.824

blend 5.04 · genu 2.57
invWER 0.80 (WER 0.20)

bottom

reward 0.201

blend 0.45 · genu 0.70
invWER 0.91 (WER 0.09)

reward 0.117

blend 0.27 · genu 0.30
invWER 0.91 (WER 0.09)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 1.212

blend 8.52 · genu 2.76
invWER 0.89 (WER 0.11)

reward 0.739

blend 6.01 · genu 0.53
invWER 1.00 (WER 0.00)

reward 0.610

blend 5.04 · genu 1.24
invWER 0.80 (WER 0.20)

bottom

reward 0.122

blend 0.45 · genu 0.24
invWER 0.91 (WER 0.09)

reward 0.097

blend 0.27 · genu 0.18
invWER 0.91 (WER 0.09)

src_d_12 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: You took credit for my work in front of everybody. I really trusted you. That was my project. But whatever, it's fine. Enjoy the praise. Now I finally see your true colors.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.097

blend 8.49 · genu 1.67
invWER 0.91 (WER 0.09)

reward 0.499

blend 2.82 · genu 1.29
invWER 0.88 (WER 0.12)

reward 0.462

blend 7.13 · genu 1.39
invWER 0.45 (WER 0.55)

bottom

reward 0.222

blend 5.91 · genu 2.20
invWER 0.21 (WER 0.79)

reward 0.164

blend 6.00 · genu 1.37
invWER 0.18 (WER 0.82)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 1.019

blend 8.49 · genu 1.31
invWER 0.91 (WER 0.09)

reward 0.435

blend 7.13 · genu 1.16
invWER 0.45 (WER 0.55)

reward 0.391

blend 2.82 · genu 0.67
invWER 0.88 (WER 0.12)

bottom

reward 0.185

blend 5.91 · genu 1.37
invWER 0.21 (WER 0.79)

reward 0.146

blend 6.00 · genu 0.91
invWER 0.18 (WER 0.82)

src_c_11 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Mom, please give my hand a squeeze if you can hear me. I'm right beside you, okay? Right here. Oh, I see it now. That glow in your eyes. You're already at peace. And I'll be alright.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.442

blend 4.58 · genu 1.26
invWER 0.60 (WER 0.40)

reward 0.435

blend 2.89 · genu 1.22
invWER 0.78 (WER 0.23)

reward 0.397

blend 8.48 · genu 1.76
invWER 0.33 (WER 0.68)

bottom

reward 0.000

blend 1.75 · genu 1.37
invWER 0.00 (WER 1.00)

reward 0.000

blend 2.93 · genu 2.23
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.477

blend 2.89 · genu 1.65
invWER 0.78 (WER 0.23)

reward 0.421

blend 4.58 · genu 1.15
invWER 0.60 (WER 0.40)

reward 0.394

blend 8.48 · genu 1.86
invWER 0.33 (WER 0.68)

bottom

reward 0.000

blend 1.75 · genu 0.99
invWER 0.00 (WER 1.00)

reward 0.000

blend 2.93 · genu 1.18
invWER 0.00 (WER 1.00)

src_b_11 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: I put them in the ground. I'd never just leave them lying out there, honey. They're people, just like us.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.974

blend 3.95 · genu 3.10
invWER 0.95 (WER 0.04)

reward 0.938

blend 5.31 · genu 1.94
invWER 1.00 (WER 0.00)

reward 0.824

blend 5.25 · genu 2.06
invWER 0.86 (WER 0.14)

bottom

reward 0.603

blend 2.94 · genu 1.42
invWER 1.00 (WER 0.00)

reward 0.290

blend 0.12 · genu 1.49
invWER 0.86 (WER 0.14)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.813

blend 5.31 · genu 1.38
invWER 1.00 (WER 0.00)

reward 0.803

blend 3.95 · genu 2.35
invWER 0.95 (WER 0.04)

reward 0.732

blend 4.37 · genu 1.66
invWER 0.95 (WER 0.04)

bottom

reward 0.504

blend 2.59 · genu 1.30
invWER 0.95 (WER 0.04)

reward 0.371

blend 0.12 · genu 2.18
invWER 0.86 (WER 0.14)

german_src_11 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: das bedeutet, ähm, wer es bislang bei keinem streaminganbieter hören konnte oder auch keine cds besitzt, der kann jetzt einfach mal auf youtube reinhören.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.465

blend 8.57 · genu 4.12
invWER 0.88 (WER 0.12)

reward 1.342

blend 8.99 · genu 4.00
invWER 0.79 (WER 0.21)

reward 1.332

blend 9.47 · genu 3.24
invWER 0.83 (WER 0.17)

bottom

reward 0.456

blend 2.43 · genu 1.13
invWER 0.92 (WER 0.08)

reward 0.403

blend 1.82 · genu 1.15
invWER 0.92 (WER 0.08)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 1.382

blend 8.99 · genu 4.72
invWER 0.79 (WER 0.21)

reward 1.370

blend 8.57 · genu 3.90
invWER 0.88 (WER 0.12)

reward 1.303

blend 9.47 · genu 3.36
invWER 0.83 (WER 0.17)

bottom

reward 0.726

blend 2.43 · genu 2.96
invWER 0.92 (WER 0.08)

reward 0.590

blend 1.82 · genu 2.44
invWER 0.92 (WER 0.08)

src_d_13 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Take one extra mouthful... there you go, gently and steadily... yes, well done... struggling... for air... save me...

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.406

blend 6.19 · genu 0.80
invWER 0.50 (WER 0.50)

reward 0.060

blend 7.44 · genu 1.52
invWER 0.06 (WER 0.94)

reward 0.052

blend 5.90 · genu 1.60
invWER 0.06 (WER 0.94)

bottom

reward 0.000

blend 7.95 · genu 1.43
invWER 0.00 (WER 1.00)

reward 0.000

blend 2.07 · genu 1.88
invWER 0.00 (WER 2.78)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.414

blend 6.19 · genu 0.94
invWER 0.50 (WER 0.50)

reward 0.052

blend 7.44 · genu 0.82
invWER 0.06 (WER 0.94)

reward 0.043

blend 5.90 · genu 0.77
invWER 0.06 (WER 0.94)

bottom

reward 0.000

blend 7.95 · genu 0.86
invWER 0.00 (WER 1.00)

reward 0.000

blend 2.07 · genu 0.85
invWER 0.00 (WER 2.78)

src_c_12 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Honestly, you two are ridiculously adorable together. Remember getting lost on that hike and him carrying you back down? I guess my only role was taking the pictures. You truly deserve this happiness. Every smile, every morning beside him. I genuinely mean it. Here is to a lifetime. To you guys. Cheers.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.148

blend 2.85 · genu 1.65
invWER 0.23 (WER 0.77)

reward 0.119

blend 4.74 · genu 2.72
invWER 0.12 (WER 0.89)

reward 0.087

blend 1.84 · genu 1.79
invWER 0.15 (WER 0.85)

bottom

reward 0.008

blend 0.00 · genu 0.55
invWER 0.06 (WER 0.94)

reward 0.000

blend 2.67 · genu 1.10
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.103

blend 2.85 · genu 0.65
invWER 0.23 (WER 0.77)

reward 0.094

blend 4.74 · genu 1.70
invWER 0.12 (WER 0.89)

reward 0.085

blend 1.84 · genu 1.88
invWER 0.15 (WER 0.85)

bottom

reward 0.008

blend 0.00 · genu 0.51
invWER 0.06 (WER 0.94)

reward 0.000

blend 2.67 · genu 0.43
invWER 0.00 (WER 1.00)

src_b_12 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: H-hold on, no. W-what do you mean you won't get better? I-if it's just a common cold or something, it ought to be gone within a week.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.257

blend 7.51 · genu 3.16
invWER 0.91 (WER 0.09)

reward 0.897

blend 4.95 · genu 2.59
invWER 0.88 (WER 0.12)

reward 0.818

blend 4.93 · genu 1.95
invWER 0.91 (WER 0.09)

bottom

reward 0.649

blend 2.80 · genu 2.53
invWER 0.81 (WER 0.19)

reward 0.022

blend 3.47 · genu 1.60
invWER 0.03 (WER 0.97)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.982

blend 7.51 · genu 1.68
invWER 0.91 (WER 0.09)

reward 0.885

blend 4.95 · genu 2.77
invWER 0.88 (WER 0.12)

reward 0.766

blend 4.04 · genu 2.16
invWER 0.94 (WER 0.06)

bottom

reward 0.424

blend 2.80 · genu 1.14
invWER 0.81 (WER 0.19)

reward 0.017

blend 3.47 · genu 0.97
invWER 0.03 (WER 0.97)

german_src_12 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: und wer bereits weiter ist, kann sich den zwölften teil mittlerweile außerdem bei allen streaminganbietern kostenfrei anhören. gehen wir jetzt zu den übrigen neuigkeiten über.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.151

blend 7.48 · genu 2.45
invWER 0.92 (WER 0.08)

reward 0.950

blend 5.28 · genu 2.46
invWER 0.92 (WER 0.08)

reward 0.941

blend 3.54 · genu 3.33
invWER 0.92 (WER 0.08)

bottom

reward 0.254

blend 1.17 · genu 0.68
invWER 0.88 (WER 0.12)

reward 0.093

blend 0.41 · genu 0.11
invWER 0.88 (WER 0.12)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.904

blend 7.48 · genu 1.10
invWER 0.92 (WER 0.08)

reward 0.733

blend 5.29 · genu 0.92
invWER 1.00 (WER 0.00)

reward 0.658

blend 5.14 · genu 0.90
invWER 0.92 (WER 0.08)

bottom

reward 0.189

blend 1.17 · genu 0.29
invWER 0.88 (WER 0.12)

reward 0.114

blend 0.41 · genu 0.23
invWER 0.88 (WER 0.12)

src_d_14 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: You haven't left... I was convinced I made you up. Wait, please don't go. I want to find out what happens if I step a little closer.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.056

blend 7.03 · genu 1.85
invWER 0.97 (WER 0.03)

reward 0.945

blend 7.48 · genu 1.62
invWER 0.86 (WER 0.14)

reward 0.936

blend 7.39 · genu 1.85
invWER 0.83 (WER 0.17)

bottom

reward 0.045

blend 3.18 · genu 1.54
invWER 0.07 (WER 0.93)

reward 0.028

blend 0.32 · genu 1.02
invWER 0.10 (WER 0.90)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.916

blend 7.03 · genu 1.17
invWER 0.97 (WER 0.03)

reward 0.875

blend 7.48 · genu 1.29
invWER 0.86 (WER 0.14)

reward 0.838

blend 7.39 · genu 1.33
invWER 0.83 (WER 0.17)

bottom

reward 0.037

blend 3.18 · genu 0.97
invWER 0.07 (WER 0.93)

reward 0.036

blend 0.32 · genu 1.58
invWER 0.10 (WER 0.90)

src_c_13 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: The water is perfectly calm today. I feel the sunlight on my closed eyes. My dad used to say to just relax your arms completely, as if you're trusting the heavens. But I need to speak it aloud. For my daughter's laughter, and for every summer I missed. I'm so overwhelmed I can hardly catch my breath.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.662

blend 3.24 · genu 1.69
invWER 0.97 (WER 0.03)

reward 0.515

blend 3.66 · genu 1.93
invWER 0.67 (WER 0.33)

reward 0.471

blend 3.03 · genu 1.92
invWER 0.67 (WER 0.33)

bottom

reward 0.198

blend 3.31 · genu 1.53
invWER 0.30 (WER 0.70)

reward 0.041

blend 4.64 · genu 1.71
invWER 0.05 (WER 0.95)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.567

blend 3.24 · genu 1.26
invWER 0.97 (WER 0.03)

reward 0.416

blend 3.66 · genu 1.23
invWER 0.67 (WER 0.33)

reward 0.346

blend 3.03 · genu 0.99
invWER 0.67 (WER 0.33)

bottom

reward 0.178

blend 4.60 · genu 1.51
invWER 0.23 (WER 0.77)

reward 0.040

blend 4.64 · genu 1.66
invWER 0.05 (WER 0.95)

src_b_13 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: There are times I forget I'm a vampire. I just feel completely ordinary.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.999

blend 5.66 · genu 2.08
invWER 1.00 (WER 0.00)

reward 0.973

blend 6.53 · genu 1.87
invWER 0.93 (WER 0.07)

reward 0.900

blend 2.94 · genu 3.00
invWER 1.00 (WER 0.00)

bottom

reward 0.590

blend 1.12 · genu 2.31
invWER 1.00 (WER 0.00)

reward 0.435

blend 4.17 · genu 1.15
invWER 0.64 (WER 0.36)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.877

blend 6.53 · genu 1.43
invWER 0.93 (WER 0.07)

reward 0.856

blend 5.66 · genu 1.43
invWER 1.00 (WER 0.00)

reward 0.704

blend 4.48 · genu 1.23
invWER 1.00 (WER 0.00)

bottom

reward 0.442

blend 4.17 · genu 1.31
invWER 0.64 (WER 0.36)

reward 0.328

blend 1.12 · genu 0.99
invWER 1.00 (WER 0.00)

german_src_13 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: die nächste welle des collectors club, insgesamt bereits die zehnte, ist nun beendet und die pdfs sind erschienen.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.252

blend 6.77 · genu 3.22
invWER 0.94 (WER 0.06)

reward 0.901

blend 4.90 · genu 2.24
invWER 0.94 (WER 0.06)

reward 0.590

blend 2.19 · genu 1.74
invWER 1.00 (WER 0.00)

bottom

reward 0.305

blend 0.92 · genu 1.00
invWER 0.94 (WER 0.06)

reward 0.242

blend 0.59 · genu 0.74
invWER 1.00 (WER 0.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.858

blend 6.77 · genu 1.08
invWER 0.94 (WER 0.06)

reward 0.544

blend 4.90 · genu 0.22
invWER 0.94 (WER 0.06)

reward 0.364

blend 1.65 · genu 1.32
invWER 0.83 (WER 0.17)

bottom

reward 0.124

blend 0.92 · genu -0.06
invWER 0.94 (WER 0.06)

reward 0.093

blend 0.59 · genu -0.09
invWER 1.00 (WER 0.00)

src_d_15 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Just a bit further... almost there, honey. Daddy's tough, right? Just... give me a minute here. That was the weekend we painted your bedroom. She chose the shade. We'd better... just put it down.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.556

blend 6.98 · genu 2.33
invWER 0.47 (WER 0.53)

reward 0.551

blend 7.96 · genu 2.13
invWER 0.44 (WER 0.56)

reward 0.226

blend 5.55 · genu 1.63
invWER 0.25 (WER 0.75)

bottom

reward 0.044

blend 3.49 · genu 2.12
invWER 0.06 (WER 0.94)

reward 0.000

blend 6.19 · genu 1.84
invWER 0.00 (WER 2.08)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.611

blend 6.98 · genu 3.24
invWER 0.47 (WER 0.53)

reward 0.525

blend 7.96 · genu 2.00
invWER 0.44 (WER 0.56)

reward 0.223

blend 5.55 · genu 1.71
invWER 0.25 (WER 0.75)

bottom

reward 0.038

blend 3.49 · genu 1.66
invWER 0.06 (WER 0.94)

reward 0.000

blend 6.19 · genu 2.30
invWER 0.00 (WER 2.08)

src_c_14 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Hey, did you bring your favorite snacks? I know, dumb question. I never told you how much you mattered. You did. You still do. I'm incredibly proud. And totally heartbroken.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.332

blend 5.85 · genu 2.14
invWER 0.32 (WER 0.68)

reward 0.315

blend 2.90 · genu 1.48
invWER 0.52 (WER 0.48)

reward 0.307

blend 3.66 · genu 0.99
invWER 0.52 (WER 0.48)

bottom

reward 0.088

blend 4.71 · genu 2.11
invWER 0.10 (WER 0.90)

reward 0.004

blend 0.00 · genu 0.35
invWER 0.03 (WER 0.97)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.346

blend 2.90 · genu 1.96
invWER 0.52 (WER 0.48)

reward 0.306

blend 3.66 · genu 1.05
invWER 0.52 (WER 0.48)

reward 0.296

blend 5.85 · genu 1.69
invWER 0.32 (WER 0.68)

bottom

reward 0.111

blend 2.74 · genu 0.64
invWER 0.26 (WER 0.74)

reward 0.007

blend 0.00 · genu 1.07
invWER 0.03 (WER 0.97)

src_b_14 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Obviously you're cute. We've been messaging back and forth about video games and anime for a long time.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.981

blend 2.68 · genu 3.57
invWER 1.00 (WER 0.00)

reward 0.970

blend 4.92 · genu 2.89
invWER 0.90 (WER 0.10)

reward 0.963

blend 4.77 · genu 2.92
invWER 0.90 (WER 0.10)

bottom

reward 0.571

blend 2.58 · genu 1.44
invWER 1.00 (WER 0.00)

reward 0.436

blend 1.90 · genu 1.08
invWER 1.00 (WER 0.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.970

blend 4.92 · genu 3.18
invWER 0.90 (WER 0.10)

reward 0.891

blend 4.77 · genu 2.74
invWER 0.90 (WER 0.10)

reward 0.777

blend 2.99 · genu 2.78
invWER 0.95 (WER 0.05)

bottom

reward 0.443

blend 2.57 · genu 1.10
invWER 0.90 (WER 0.10)

reward 0.331

blend 1.90 · genu 0.54
invWER 1.00 (WER 0.00)

german_src_14 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: am 13.07., also einen tag nachdem die welle vorüber war,

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.115

blend 5.76 · genu 3.23
invWER 0.91 (WER 0.09)

reward 0.945

blend 4.55 · genu 2.37
invWER 1.00 (WER 0.00)

reward 0.873

blend 3.87 · genu 2.36
invWER 1.00 (WER 0.00)

bottom

reward 0.426

blend 2.44 · genu 2.04
invWER 0.64 (WER 0.36)

reward 0.271

blend 1.43 · genu 1.28
invWER 0.64 (WER 0.36)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 1.157

blend 4.55 · genu 3.87
invWER 1.00 (WER 0.00)

reward 1.089

blend 5.76 · genu 3.39
invWER 0.91 (WER 0.09)

reward 0.872

blend 3.87 · genu 2.58
invWER 1.00 (WER 0.00)

bottom

reward 0.493

blend 2.44 · genu 2.86
invWER 0.64 (WER 0.36)

reward 0.454

blend 1.43 · genu 3.09
invWER 0.64 (WER 0.36)

src_d_16 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Mmm, that felt great. Yeah... really great. Oh my god... oh my god, wait... who are you? No, no, no... please don't...

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.617

blend 6.18 · genu 1.88
invWER 0.61 (WER 0.39)

reward 0.147

blend 7.98 · genu 1.54
invWER 0.13 (WER 0.87)

reward 0.122

blend 6.17 · genu 1.48
invWER 0.13 (WER 0.87)

bottom

reward 0.000

blend 4.56 · genu 1.48
invWER 0.00 (WER 1.00)

reward 0.000

blend 5.92 · genu 1.47
invWER 0.00 (WER 1.30)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.493

blend 6.18 · genu 0.85
invWER 0.61 (WER 0.39)

reward 0.147

blend 7.98 · genu 1.64
invWER 0.13 (WER 0.87)

reward 0.119

blend 6.17 · genu 1.46
invWER 0.13 (WER 0.87)

bottom

reward 0.000

blend 4.56 · genu 0.97
invWER 0.00 (WER 1.00)

reward 0.000

blend 5.92 · genu 0.53
invWER 0.00 (WER 1.30)

src_c_15 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Do you think I showed up to fight? I just came to say goodbye. Nothing more. No arguing. It's over. God, you're standing exactly like you did that night in the hospital. I hate that I still recognize the way you hold yourself.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.315

blend 6.67 · genu 1.60
invWER 0.31 (WER 0.69)

reward 0.312

blend 6.39 · genu 1.71
invWER 0.31 (WER 0.69)

reward 0.303

blend 9.10 · genu 2.17
invWER 0.22 (WER 0.78)

bottom

reward 0.050

blend 6.11 · genu 2.55
invWER 0.04 (WER 0.96)

reward 0.034

blend 2.80 · genu 0.96
invWER 0.07 (WER 0.93)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.315

blend 6.39 · genu 1.93
invWER 0.31 (WER 0.69)

reward 0.301

blend 6.67 · genu 1.49
invWER 0.31 (WER 0.69)

reward 0.268

blend 9.10 · genu 1.48
invWER 0.22 (WER 0.78)

bottom

reward 0.043

blend 6.11 · genu 1.82
invWER 0.04 (WER 0.96)

reward 0.041

blend 2.80 · genu 1.71
invWER 0.07 (WER 0.93)

src_b_15 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Awww. Well, I know how much you love it when I wriggle and struggle all bundled up.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.008

blend 6.72 · genu 1.89
invWER 0.94 (WER 0.06)

reward 0.880

blend 4.40 · genu 2.41
invWER 0.94 (WER 0.06)

reward 0.876

blend 5.21 · genu 1.95
invWER 0.94 (WER 0.06)

bottom

reward 0.631

blend 4.18 · genu 2.30
invWER 0.71 (WER 0.29)

reward 0.000

blend 5.29 · genu 1.93
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.901

blend 6.72 · genu 1.40
invWER 0.94 (WER 0.06)

reward 0.742

blend 6.20 · genu 0.71
invWER 0.94 (WER 0.06)

reward 0.709

blend 5.08 · genu 1.17
invWER 0.94 (WER 0.06)

bottom

reward 0.429

blend 3.26 · genu 0.48
invWER 0.94 (WER 0.06)

reward 0.000

blend 5.29 · genu 1.42
invWER 0.00 (WER 1.00)

german_src_15 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: außerdem haben wir den fluch der hornechse hier. es handelt sich um den sechsten teil der rahmenkriegkampagne. achtung, spoilergefahr! wir betrachten jetzt gleich die rückseite und blättern auch kurz hinein. der band ist für 14,95 zu haben. man erkennt zudem lebendige geschichte. es gibt 4 von 4 punkten.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.716

blend 4.85 · genu 2.24
invWER 0.76 (WER 0.24)

reward 0.672

blend 3.38 · genu 2.05
invWER 0.88 (WER 0.12)

reward 0.658

blend 3.62 · genu 1.93
invWER 0.86 (WER 0.14)

bottom

reward 0.451

blend 2.34 · genu 1.39
invWER 0.84 (WER 0.16)

reward 0.365

blend 2.43 · genu 0.92
invWER 0.80 (WER 0.20)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.644

blend 4.85 · genu 1.89
invWER 0.76 (WER 0.24)

reward 0.546

blend 3.62 · genu 1.34
invWER 0.86 (WER 0.14)

reward 0.541

blend 3.51 · genu 1.20
invWER 0.90 (WER 0.10)

bottom

reward 0.274

blend 2.34 · genu 0.26
invWER 0.84 (WER 0.16)

reward 0.240

blend 2.43 · genu 0.06
invWER 0.80 (WER 0.20)

src_d_17 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Yeah, this river is a worn-out old beast. But deep down, she has a gentle soul. You just need to... give her some respect. You miserable bastard! Let go! Get off my leg!

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.149

blend 6.75 · genu 1.57
invWER 0.15 (WER 0.85)

reward 0.132

blend 3.79 · genu 1.17
invWER 0.21 (WER 0.79)

reward 0.119

blend 6.68 · genu 1.58
invWER 0.12 (WER 0.88)

bottom

reward 0.000

blend 0.56 · genu 0.73
invWER 0.00 (WER 1.00)

reward 0.000

blend 0.00 · genu 0.79
invWER 0.00 (WER 1.44)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.148

blend 6.75 · genu 1.69
invWER 0.15 (WER 0.85)

reward 0.140

blend 4.22 · genu 2.84
invWER 0.15 (WER 0.85)

reward 0.137

blend 3.79 · genu 1.40
invWER 0.21 (WER 0.79)

bottom

reward 0.000

blend 0.56 · genu -0.23
invWER 0.00 (WER 1.00)

reward 0.000

blend 0.00 · genu 1.09
invWER 0.00 (WER 1.44)

src_c_16 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Did you really think I came to pick a fight? I just wanted to say my goodbyes. Nothing else. No drama. I'm just finished. God, your posture is exactly like that knight in the emergency room. It bothers me that I still recognize the way you hold yourself.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.389

blend 2.25 · genu 1.32
invWER 0.76 (WER 0.24)

reward 0.306

blend 5.04 · genu 1.78
invWER 0.35 (WER 0.65)

reward 0.156

blend 0.60 · genu 1.15
invWER 0.49 (WER 0.51)

bottom

reward 0.000

blend 2.45 · genu 1.47
invWER 0.00 (WER 1.00)

reward 0.000

blend 1.73 · genu 1.14
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.449

blend 2.25 · genu 1.90
invWER 0.76 (WER 0.24)

reward 0.300

blend 5.04 · genu 1.85
invWER 0.35 (WER 0.65)

reward 0.215

blend 0.60 · genu 1.96
invWER 0.49 (WER 0.51)

bottom

reward 0.000

blend 2.45 · genu 1.90
invWER 0.00 (WER 1.00)

reward 0.000

blend 1.73 · genu 1.57
invWER 0.00 (WER 1.00)

src_b_16 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Sweetie, that was incredible. I can feel your blood inside me.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.213

blend 7.98 · genu 1.98
invWER 1.00 (WER 0.00)

reward 1.130

blend 7.76 · genu 1.66
invWER 1.00 (WER 0.00)

reward 1.110

blend 7.90 · genu 1.48
invWER 1.00 (WER 0.00)

bottom

reward 0.746

blend 4.32 · genu 1.44
invWER 1.00 (WER 0.00)

reward 0.713

blend 5.04 · genu 3.06
invWER 0.64 (WER 0.36)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.932

blend 7.98 · genu 0.50
invWER 1.00 (WER 0.00)

reward 0.922

blend 7.76 · genu 0.58
invWER 1.00 (WER 0.00)

reward 0.898

blend 7.71 · genu 0.46
invWER 1.00 (WER 0.00)

bottom

reward 0.334

blend 2.70 · genu 0.09
invWER 1.00 (WER 0.00)

reward 0.300

blend 2.33 · genu 0.11
invWER 1.00 (WER 0.00)

german_src_16 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: ähm, es ist von armin abel und david schmidt. schauen wir kurz ins inhaltsverzeichnis. wer hier gleich den ersten fehler entdeckt, ähm, bitte melden.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.537

blend 10.00 · genu 4.26
invWER 0.83 (WER 0.17)

reward 1.458

blend 10.00 · genu 3.76
invWER 0.83 (WER 0.17)

reward 1.327

blend 9.38 · genu 3.70
invWER 0.79 (WER 0.21)

bottom

reward 1.144

blend 7.28 · genu 2.54
invWER 0.92 (WER 0.08)

reward 1.142

blend 7.05 · genu 3.69
invWER 0.79 (WER 0.21)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 1.382

blend 10.00 · genu 3.61
invWER 0.83 (WER 0.17)

reward 1.359

blend 8.96 · genu 3.60
invWER 0.88 (WER 0.12)

reward 1.329

blend 10.00 · genu 3.23
invWER 0.83 (WER 0.17)

bottom

reward 1.193

blend 8.78 · genu 3.43
invWER 0.79 (WER 0.21)

reward 1.019

blend 7.54 · genu 2.14
invWER 0.88 (WER 0.12)

src_d_18 Source D 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: I'll kill anyone who lays another finger on you.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.953

blend 6.02 · genu 1.64
invWER 1.00 (WER 0.00)

reward 0.896

blend 5.85 · genu 1.42
invWER 1.00 (WER 0.00)

reward 0.808

blend 5.34 · genu 1.71
invWER 0.90 (WER 0.10)

bottom

reward 0.188

blend 3.65 · genu 1.16
invWER 0.30 (WER 0.70)

reward 0.000

blend 0.48 · genu 0.19
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.899

blend 6.02 · genu 1.47
invWER 1.00 (WER 0.00)

reward 0.829

blend 5.85 · genu 1.15
invWER 1.00 (WER 0.00)

reward 0.699

blend 5.34 · genu 1.15
invWER 0.90 (WER 0.10)

bottom

reward 0.166

blend 3.65 · genu 0.83
invWER 0.30 (WER 0.70)

reward 0.000

blend 0.48 · genu 0.58
invWER 0.00 (WER 1.00)

src_c_17 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: There's something I have to confess. I've been hiding the truth. It involves our relationship. It's about my past actions. I am deeply in love with you. Loving you this intensely is tearing me apart. I was terrified of you walking away. I've been dishonest. Constantly.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.266

blend 6.90 · genu 2.00
invWER 0.24 (WER 0.76)

reward 0.228

blend 6.42 · genu 1.87
invWER 0.22 (WER 0.78)

reward 0.224

blend 5.08 · genu 2.04
invWER 0.24 (WER 0.76)

bottom

reward 0.000

blend 4.73 · genu 2.21
invWER 0.00 (WER 1.00)

reward 0.000

blend 1.41 · genu 1.44
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.222

blend 6.90 · genu 1.10
invWER 0.24 (WER 0.76)

reward 0.191

blend 6.42 · genu 1.05
invWER 0.22 (WER 0.78)

reward 0.177

blend 5.08 · genu 1.07
invWER 0.24 (WER 0.76)

bottom

reward 0.000

blend 4.73 · genu 1.16
invWER 0.00 (WER 1.00)

reward 0.000

blend 1.41 · genu 0.27
invWER 0.00 (WER 1.00)

src_b_17 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Or you sit beside me rather than opposite me. Or you'd like me to message you when I arrive at and leave work?

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.645

blend 2.86 · genu 1.68
invWER 1.00 (WER 0.00)

reward 0.535

blend 0.74 · genu 2.22
invWER 1.00 (WER 0.00)

reward 0.489

blend 3.22 · genu 1.03
invWER 0.88 (WER 0.12)

bottom

reward 0.285

blend 1.11 · genu 0.69
invWER 1.00 (WER 0.00)

reward 0.222

blend 1.10 · genu 0.37
invWER 1.00 (WER 0.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.527

blend 2.86 · genu 1.14
invWER 1.00 (WER 0.00)

reward 0.410

blend 3.22 · genu 0.58
invWER 0.88 (WER 0.12)

reward 0.362

blend 1.47 · genu 0.99
invWER 1.00 (WER 0.00)

bottom

reward 0.206

blend 1.10 · genu 0.28
invWER 1.00 (WER 0.00)

reward 0.147

blend 0.50 · genu 0.29
invWER 1.00 (WER 0.00)

german_src_17 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: und ähm, außerdem nochmal, ja, eine beispielseite aus dem szenario. ich möchte jetzt aber nicht zu viel spoilern. daher, ähm, höre ich an dieser stelle auf.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.717

blend 9.09 · genu 4.07
invWER 1.00 (WER 0.00)

reward 1.653

blend 10.00 · genu 3.98
invWER 0.92 (WER 0.08)

reward 1.630

blend 8.99 · genu 4.38
invWER 0.92 (WER 0.08)

bottom

reward 1.310

blend 9.92 · genu 3.56
invWER 0.77 (WER 0.23)

reward 0.879

blend 2.75 · genu 3.60
invWER 0.88 (WER 0.12)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 1.700

blend 9.09 · genu 4.39
invWER 1.00 (WER 0.00)

reward 1.571

blend 8.99 · genu 4.47
invWER 0.92 (WER 0.08)

reward 1.561

blend 10.00 · genu 3.80
invWER 0.92 (WER 0.08)

bottom

reward 1.315

blend 9.92 · genu 3.96
invWER 0.77 (WER 0.23)

reward 0.813

blend 2.75 · genu 3.52
invWER 0.88 (WER 0.12)

src_d_19 Source D 🇺🇸 EN <5 variants

Reference (original speaker)

Paraphrased text spoken: She even brought out the fancy stationery, which is exactly like her. Notice how she wrote with a fountain pen. She was aware. She realized this was her final farewell. Thanks, Mom. I am furious that you left me. Thank you.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.501

blend 3.92 · genu 2.05
invWER 0.61 (WER 0.39)

reward 0.076

blend 4.53 · genu 0.69
invWER 0.12 (WER 0.88)

reward 0.000

blend 6.62 · genu 1.79
invWER 0.00 (WER 1.00)

bottom

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.393

blend 3.92 · genu 1.21
invWER 0.61 (WER 0.39)

reward 0.099

blend 4.53 · genu 1.85
invWER 0.12 (WER 0.88)

reward 0.000

blend 6.62 · genu 3.47
invWER 0.00 (WER 1.00)

bottom

src_c_18 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: There is something I have to confess. I've been hiding the truth. It involves our relationship and my actions. I love you. I love you so deeply it's tearing me apart. I was terrified you would leave me. I've been dishonest. Day after day.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.591

blend 3.59 · genu 1.96
invWER 0.77 (WER 0.23)

reward 0.425

blend 1.47 · genu 2.03
invWER 0.74 (WER 0.26)

reward 0.396

blend 1.84 · genu 2.09
invWER 0.64 (WER 0.36)

bottom

reward 0.155

blend 3.52 · genu 1.77
invWER 0.21 (WER 0.79)

reward 0.069

blend 2.63 · genu 1.25
invWER 0.13 (WER 0.87)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.503

blend 3.59 · genu 1.47
invWER 0.77 (WER 0.23)

reward 0.320

blend 2.42 · genu 1.57
invWER 0.57 (WER 0.43)

reward 0.291

blend 1.47 · genu 1.15
invWER 0.74 (WER 0.26)

bottom

reward 0.151

blend 3.52 · genu 1.82
invWER 0.21 (WER 0.79)

reward 0.065

blend 2.63 · genu 1.17
invWER 0.13 (WER 0.87)

src_b_18 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: Since I no longer have to wait for a special occasion to pamper you, I can treat you around the clock. Bringing you joy is my true love language.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.554

blend 2.44 · genu 1.52
invWER 0.97 (WER 0.03)

reward 0.540

blend 1.75 · genu 1.71
invWER 1.00 (WER 0.00)

reward 0.448

blend 0.28 · genu 2.01
invWER 1.00 (WER 0.00)

bottom

reward 0.179

blend 0.30 · genu 0.56
invWER 1.00 (WER 0.00)

reward 0.162

blend 0.00 · genu 0.63
invWER 1.00 (WER 0.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.378

blend 2.44 · genu 0.58
invWER 0.97 (WER 0.03)

reward 0.298

blend 1.48 · genu 0.60
invWER 1.00 (WER 0.00)

reward 0.264

blend 1.75 · genu 0.23
invWER 1.00 (WER 0.00)

bottom

reward 0.082

blend 0.00 · genu 0.20
invWER 1.00 (WER 0.00)

reward 0.080

blend 0.00 · genu 0.19
invWER 1.00 (WER 0.00)

german_src_18 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: außerdem haben wir noch das spielkartenset zum aventurischen animatorium. es erscheint ebenfalls am 28.10. für 24,95.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.925

blend 5.13 · genu 5.09
invWER 0.61 (WER 0.39)

reward 0.672

blend 4.23 · genu 2.88
invWER 0.67 (WER 0.33)

reward 0.664

blend 3.03 · genu 1.90
invWER 0.94 (WER 0.06)

bottom

reward 0.306

blend 1.03 · genu 1.32
invWER 0.78 (WER 0.22)

reward 0.167

blend 0.08 · genu 0.73
invWER 0.89 (WER 0.11)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.501

blend 4.23 · genu 1.66
invWER 0.67 (WER 0.33)

reward 0.475

blend 2.94 · genu 1.59
invWER 0.78 (WER 0.22)

reward 0.469

blend 5.07 · genu 1.70
invWER 0.56 (WER 0.44)

bottom

reward 0.212

blend 1.03 · genu 0.72
invWER 0.78 (WER 0.22)

reward 0.102

blend 0.08 · genu 0.34
invWER 0.89 (WER 0.11)

src_c_19 Source C 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: I wish I had apologized. I threw away so much time being angry. All the chances I missed. But then I recall the rain on the warm asphalt. The way she laughed. That single flawless October morning.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 0.436

blend 5.19 · genu 1.78
invWER 0.49 (WER 0.51)

reward 0.420

blend 2.49 · genu 0.68
invWER 1.00 (WER 0.00)

reward 0.402

blend 2.09 · genu 1.49
invWER 0.76 (WER 0.24)

bottom

reward 0.014

blend 1.30 · genu 1.83
invWER 0.03 (WER 0.97)

reward 0.000

blend 2.85 · genu 1.74
invWER 0.00 (WER 1.00)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.537

blend 2.49 · genu 1.41
invWER 1.00 (WER 0.00)

reward 0.409

blend 5.19 · genu 1.61
invWER 0.49 (WER 0.51)

reward 0.402

blend 2.09 · genu 1.61
invWER 0.76 (WER 0.24)

bottom

reward 0.011

blend 1.30 · genu 1.26
invWER 0.03 (WER 0.97)

reward 0.000

blend 2.85 · genu 1.20
invWER 0.00 (WER 1.00)

src_b_19 Source B 🇺🇸 EN

Reference (original speaker)

Paraphrased text spoken: I love looking after you. I want you always... Yeah, I'll be your sexy little sex fiend-

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2

reward 1.006

blend 3.60 · genu 3.88
invWER 0.89 (WER 0.11)

reward 0.993

blend 3.91 · genu 3.28
invWER 0.94 (WER 0.06)

reward 0.945

blend 3.53 · genu 2.92
invWER 1.00 (WER 0.00)

bottom

reward 0.200

blend 3.43 · genu 2.74
invWER 0.22 (WER 0.78)

reward 0.124

blend 1.91 · genu 2.71
invWER 0.17 (WER 0.83)

VoiceCLAP-large-v2 (3584-d) combined reward — top-3 / bottom-2

reward 0.723

blend 3.60 · genu 2.40
invWER 0.89 (WER 0.11)

reward 0.698

blend 3.91 · genu 1.77
invWER 0.94 (WER 0.06)

reward 0.623

blend 3.53 · genu 1.31
invWER 1.00 (WER 0.00)

bottom

reward 0.141

blend 3.43 · genu 1.44
invWER 0.22 (WER 0.78)

reward 0.066

blend 1.91 · genu 0.94
invWER 0.17 (WER 0.83)

german_src_19 German source 🇩🇪 DE

Reference (original speaker)

Paraphrased text spoken: und, äh, auch hier können wir vorab keinen blick riskieren, weil es ein produkt ohne feedbackphase ist. das heißt, die pdfs gibt es dann wirklich erst mit dem erscheinen des finalen sets. die vorbesteller hatten, äh, fälschlicherweise eine mail bekommen, in der das aventurische animatorium auch schon zum herunterladen verfügbar war.

VoiceCLAP-commercial (768-d) combined reward — top-3 / bottom-2