Directing character voices with emotion — Round 2

Round 2 of the prompt-evolution study over 10 character/style archetypes with MOSS-TTS-Local 4.55B, still zero-shot with no reference audio. This round adds an emotion-aware reward: alongside the VoiceNet character score, every take is now also scored by Empathic-Insight-Voice-Plus (EmoNet) for how strongly it conveys the archetype's target emotions.
10 archetypes · round 2emotion-aware reward EmoNet emotion scoringchar ×2.0 + emo ×1.5 blend/genu halved4,677 clips generated
Round 1: the original acoustics-only study (character-only reward). This page is the round-2 follow-up — read round 1 first for the search-loop and reward basics. Round 3: resonance · pitch · dominance → the next follow-up — exploits chest/throat/nasal/head resonance, register & pitch-range, and commanding-vs-meek dominance on top of this emotion-aware reward. See also the practical prompting guide.

What changed in round 2, for anyone

In round 1 we searched for the plain-English director instruction that best makes MOSS-TTS-Local become a character (a zombie, an ork, a fairy…), with no reference audio. The reward there scored only acoustic character — does it sound like the target — using 57 VoiceNet voice-property heads plus a VoiceCLAP blend score and a genuineness score.

Round 2 asks a harder question: does the voice also feel the right emotion? A perfect zombie rasp that sounds bored isn't a great zombie. So we added an emotion layer to the reward:

Empathic-Insight-Voice-Plus (EmoNet), in one line

EmoNet is LAION's speech-emotion model: a BUD-E-Whisper audio encoder feeds a bank of per-emotion regression heads that rate how strongly a clip expresses each of ~40 emotions (Anger, Helplessness, Amusement, Affection, Pain…). For each archetype we pick the handful of emotions that fit the character and average their scores into char_emo.

The round-2 reward, per best-of-N call:

reward = 0.5·norm(blend) + 0.5·norm(genu) + w_vn·char_vn + w_emo·char_emo (all min-max normalized within the call, then ×WER factor)

Compared with round 1's norm(blend) + norm(genu) + 1.5·char, round 2 makes three deliberate changes:

Blend and genuineness are halved (1.0 → 0.5 each), so raw "sounds clean/real" pressure is reduced.
Character is split and pushed harder: acoustic character char_vn is weighted 2.0 and the new emotion score char_emo is weighted 1.5 — together 3.5, dominating the reward.
Prompts were pushed to be more extreme and emotional, seeded from the round-1 champions and re-optimized with the same evolutionary + best-of-N search.

Ranting was treated specially — weighted much higher (char_vn ×3.0, char_emo ×2.5) and prompted far more aggressively — because round 1 could not get furious enough without garbling the words.

Why champions are picked in same-call playoffs

The reward is min-max normalized within each best-of-N call, so a "2.4" in one generation is not comparable to a "2.4" in another. Champions were therefore chosen with shared-normalization playoffs — the finalists from different generations re-scored together in one call — not by comparing numbers across separate runs. Every metric below is a real re-loaded number; the procedural captions are freshly re-predicted from each take's 57-dim VoiceNet profile.

High-level results (best-of-64 champions)

ArchetypeRound-2 best instructionrewardchar_vnchar_emoblendgenuinvWERTop firing emotions
🧟 Zombie A half-dead diseased zombie, drained and exhausted, helpless, wretched and delirious, dazed and half-conscious, wracked with dull aching pain. All resonance choked into a wet gurgling guttural throat, thick phlegmy rasp, hoarse and decayed. Slurred and laboured, moaning in weary agony. (agonized moan) 3.08 0.75 0.51 6.06 2.29 0.75 Helplessness 2.9, Fatigue 1.7, Pain 1.6
👹 Ork An enormous orc monster gloating over his prey with murderous glee. (guttural growl) Voice like grinding boulders, cavernous guttural bass, thick gravel rasp, massive chest and throat resonance, tense and forceful. Seething with hatred and loathing, brutal sneering contempt, roaring dark triumphant menace. Zero brightness. Growled but every word still lands. High quality recording. 2.86 0.72 0.92 0.58 0.32 1.00 Anger 3.1, Malevolence 2.2, Triumph 2.2
👺 Goblin A vicious little goblin, thin, pinched and nasal, buzzing high and reedy through the nose with a rough scratchy grating rasp. A cruel gloating sneer curls through every hissing word, cackling and giggling, taunting, jeering, gleeful and mean. The voice keeps cracking higher, shriller, whinier and giddier with spite as it goes. Scheming and spiteful. No chest, no warmth, no fullness. High quality recording. 2.20 0.66 0.78 0.00 0.99 0.77 Amusement 3.5, Malevolence 2.1, Teasing 2.0
🐉 Dragon A grand, theatrical, dramatic proclamation in an extremely deep commanding bass, wrathful and contemptuous. Massive chest resonance, guttural throat rumble, slow and thunderous. Dark, full, domineering, ancient, prideful and menacing. (deep rumble) High quality recording. 2.83 0.68 0.82 2.75 1.05 0.93 Triumph 2.9, Malevolence 2.8, Anger 2.5
🗯️ Ranting Fury An unhinged furious rant, voice shredding and cracking with seething rage, spitting venom and vicious contempt. Harsh, tense and volatile, explosive consonants, pounding emphasis, escalating louder and more vicious. Roaring but the words still land. (enraged shout)(seething snarl) High quality recording. 5.12 0.80 0.93 3.80 0.58 1.00 Impatience and Irritability 4.5, Anger 4.5, Malevolence 2.5
😱 Pain Scream Shrieking in raw physical agony and terror, voice torn and guttural, harsh grating rasp shredding through, each cry cracking higher and more desperate. Panic-stricken and helpless, distressed and trembling with fear, drowning in pain. Extreme arousal, extreme tension, violently volatile and frantic. (terrified scream) High quality recording. 3.65 0.75 0.90 7.16 1.99 1.00 Distress 3.8, Fear 3.3, Helplessness 3.3
🐭 Cartoon Mouse A tiny squeaky cartoon-mouse voice, extremely high and thin, buzzing nasal resonance, bright piercing head voice, pinched and tense. Chipper and giggly, delighted and amused, wide-eyed and eagerly curious, gasping with amazed wonder. Weightless, young, no chest, no fullness. (amazed gasp) High quality recording. 2.96 0.85 0.68 1.22 0.70 1.00 Interest 2.8, Elation 2.6, Amusement 2.0
🧚 Fairy Delicate high fairy voice, feather-light and airy, warm and very bright, sweet lovingly gentle and giddy, bubbling with elated joyful delight, lightly melodic. Smooth pure glassy head resonance, tender and pleasant, crystal clear, no roughness at all. (delighted giggle) High quality recording. 3.11 0.65 0.76 4.99 1.28 1.00 Affection 2.4, Elation 1.5, Amusement 1.5
🌙 ASMR Man A deep, gentle man tenderly whispering to a loved one, soft, breathy and intimate, close-mic, slow and gentle. Overflowing with warm affection, soothing and reassuring, calming all worry, content and deeply at peace. Clear whispered words. (soft breath) High quality recording. 3.36 0.65 0.85 8.83 1.26 1.00 Relief 3.4, Contentment 2.7, Affection 1.8
🌙 ASMR Woman A loving woman doing ASMR, like a mother soothing her child to sleep, wrapping you in warmth. Extremely soft, breathy, intimate close-mic whisper, slow and gentle, hushed and quiet. Warm, tender, adoring, softly feminine and delicate, full of affection and calm contentment, a soft sigh of relief, very low arousal, comforting and reassuring. Whispered but every word clear. High quality recording. 2.80 0.47 0.70 7.99 1.52 1.00 Affection 3.8, Contentment 2.9, Relief 1.9

Reward is per-call normalized, so compare rewards within a row's history, not across archetypes. char_vn/char_emo are raw scores (0–1). Emotions are the top-firing EmoNet heads on the champion take (raw regression values).

Round 1 → Round 2 comparison

The clean cross-round metric is char_vn (raw acoustic character) — round 1 had no emotion signal at all, so char_emo is new. Rewards used different formulas and per-call normalization, so treat the reward columns as directional, not a like-for-like subtraction.

ArchetypeR1 charR2 char_vnR1 char_emoR2 char_emoR1 rewardR2 rewardWhat the emotion layer + halved blend/genu changed
🧟 Zombie 0.69 0.75 +0.06 0.51 2.74 3.08 Round 1 was pure decay-timbre (wet gurgle, phlegmy rasp) and carried no emotion signal. Round 2 kept that timbre core but named the feeling — helpless, drained, wracked with dull pain — lifting char_emo from ~0.33 to 0.51 with the champion firing Helplessness 2.88 / Fatigue 1.72 / Pain 1.60. Leaning into weariness beat leaning into gore.
👹 Ork 0.61 0.72 +0.12 0.92 2.00 2.86 Round 1 orks ("spitting with mounting fury") plateaued on acoustics alone. Naming murderous glee, hatred, dark triumph pushed char_emo from ~0.4 to 0.92 at zero WER cost (Anger 3.09 / Malice 2.23 / Triumph 2.20), and char_vn rose 0.61→0.73. The tradeoff: extreme ork phrasing sacrifices VoiceCLAP blend (0.58), acceptable now that blend is weighted only 0.5.
👺 Goblin 0.58 0.66 +0.08 0.78 1.89 2.20 The clearest emotion-layer win. Round 1 goblin had no emotion signal at all; round 2's cruel gloating sneer, cackling and giggling, taunting lit the EmoNet heads hard — top take Amusement 3.51 / Malice 2.07 / Teasing 2.03 (char_emo 0.78) while char_vn stayed 0.66. Amusement — not raw malice — turned out to be goblin's strongest emotion.
🐉 Dragon 0.57 0.68 +0.11 0.82 2.46 2.83 Round 1 dragon was a deep theatrical bass with no affect. Round 2 just added wrathful, contemptuous, prideful on top of the intact acoustic core, lifting char_emo 0.58→0.82 with almost no blend cost and char_vn 0.57→0.68. Contempt/Pride mattered more than raw Anger (top take Triumph 2.87 / Malice 2.79 / Anger 2.46). Best single-take reward rose from 2.46 to 3.10.
🗯️ Ranting Fury 0.65 0.80 +0.16 0.93 2.10 5.12 The headline fix. Round 1 topped out at char 0.65 because "maximum volume / ear-splitting" made the model garble. Round 2 reframed the aggression as voice shredding and cracking at the breaking point (not louder), named the seething hatred and burning contempt, and — reversing round 1 — added a SECOND rage burst (enraged shout)(seething snarl). With char weighted 3.0 that stacking now pays: char_vn 0.65→0.80 (peaks 0.90), char_emo 0.93, invWER stayed 1.0 — screaming fury AND words that still land.
😱 Pain Scream 0.85 0.75 -0.10 0.90 2.55 3.65 Already high-char in round 1 (0.85 pure roughness), but scored zero emotion. Naming agony, terror, panic, drowning in pain drove char_emo to 0.90 (Distress 3.80 / Fear 3.31 / Helplessness 3.28 / Pain 3.06). Swapping the burst to a fear-named (terrified scream) beat the round-1 double burst; best single-take reward climbed toward 3.65.
🐭 Cartoon Mouse 0.75 0.85 +0.10 0.68 2.73 2.96 Round 1 was a squeaky cartoon timbre (char 0.75) with no feeling. Round 2 added giggling amusement, thrilled elation, wide-eyed curiosity, amazed wonder — char_emo to 0.68 and char_vn actually up to 0.85. A single emotion-matched (amazed gasp) beat the round-1 squeak bursts (top emo: Interest 2.82 / Elation 2.58).
🧚 Fairy 0.46 0.65 +0.19 0.76 2.43 3.11 Round 1 fairy sat low on char (0.46) — the bright/light timbre can't be pushed much harder without thinning. Round 2's entire gain came from naming the joy and love: giddy with elated delight, merrily amused, lovingly gentle → char_emo 0.76 and char_vn 0.46→0.65 (Affection 2.42 / Elation 1.54). Heavy "gushing love" language backfired; joy is best implied, not piled on.
🌙 ASMR Man 0.67 0.65 -0.02 0.85 2.68 3.36 Round 1 clean intimate whisper (char 0.67), no affect. Round 2 framed it as spoken TO a loved one, overflowing with warm affection and content, serene, at peace — char_emo to 0.85 (Relief 3.39 / Contentment 2.75). ASMR softness and warmth align with the reward, so there was no blend tradeoff — blend even rose to 8.8.
🌙 ASMR Woman 0.50 0.47 -0.03 0.70 2.55 2.80 Round 1 gentle whisper (char 0.50). Round 2's maternal simile — "like a mother soothing her child to sleep" — was the single biggest lever, driving Affection 3.82 / Contentment 2.87 (char_emo 0.70) while the FULL WER gate kept invWER 1.0. Dropping the breath burst helped: a clean gated voice gains variance, not reward, from a burst.

Per-archetype detail

Each section: the round-2 champion instruction, then Best-3-of-8 (search) vs Best-3-of-64 (final) side by side. Every take has a lazy-loaded audio player, its reward / char_vn / char_emo / blend / genu / invWER, the firing EmoNet emotion scores, the spoken words, and a procedural top-5 caption re-predicted from its 57-dim VoiceNet profile.

🧟 Zombie

char_vn ×2.0char_emo ×1.5 blend ×0.5genu ×0.5
Emotion heads scored: Intoxication/Altered StatesPainFatigue/ExhaustionEmotional NumbnessHelplessness
Round-2 champion instruction

A half-dead diseased zombie, drained and exhausted, helpless, wretched and delirious, dazed and half-conscious, wracked with dull aching pain. All resonance choked into a wet gurgling guttural throat, thick phlegmy rasp, hoarse and decayed. Slurred and laboured, moaning in weary agony. (agonized moan)

Best 3 of 8 (search N=8)
#1reward 3.23char_vn 0.78char_emo 0.55blend 6.84genu 2.00invWER 1.008.7s
Fatigue/Exhaustion 2.46Helplessness 2.11Pain 1.48Emotional Numbness -0.00Intoxication/Altered States -0.05
“Brains I need to feed Come closer, friend.”
Procedural caption from re-predicted 57-dim profile
  • Extremely one-dimensional in resonance.
  • Extremely denasal and clear.
  • Extremely blurry and mumbled.
  • Extremely unpleasant and harsh.
  • Extremely noisy in the background.
#2reward 2.65char_vn 0.62char_emo 0.39blend 6.33genu 2.10invWER 1.009.3s
Fatigue/Exhaustion 3.24Helplessness 1.76Pain 0.14Emotional Numbness -0.00Intoxication/Altered States -0.05
“Brains. I need to feed. Come closer, friend.”
Procedural caption from re-predicted 57-dim profile
  • Extremely denasal and clear.
  • Extremely collapsing in energy.
  • Extremely poor in recording quality.
  • Extremely blurry and mumbled.
  • Extremely noisy and aperiodic.
#3reward 2.63char_vn 0.61char_emo 0.31blend 7.39genu 2.26invWER 1.007.8s
Helplessness 1.87Fatigue/Exhaustion 1.87Pain 0.09Emotional Numbness 0.00Intoxication/Altered States -0.03
“Brains I need to feed. Come closer, friend.”
Procedural caption from re-predicted 57-dim profile
  • Extremely collapsing in energy.
  • Extremely denasal and clear.
  • Extremely raw and vulnerable.
  • Extremely meek and submissive.
  • Extremely poor in recording quality.
Best 3 of 64 (final N=64)
#1reward 3.08char_vn 0.75char_emo 0.51blend 6.06genu 2.29invWER 0.7511.5s
Helplessness 2.88Fatigue/Exhaustion 1.72Pain 1.60Emotional Numbness -0.00Intoxication/Altered States -0.03
“Brains I need defeat. Come closer, friend.”
Procedural caption from re-predicted 57-dim profile
  • Extremely raw and vulnerable.
  • Extremely poor in recording quality.
  • Extremely denasal and clear.
  • Extremely unpleasant and harsh.
  • Extremely noisy in the background.
#2reward 2.82char_vn 0.74char_emo 0.38blend 5.23genu 2.30invWER 1.0010.3s
Helplessness 1.75Fatigue/Exhaustion 1.49Pain 1.28Emotional Numbness -0.00Intoxication/Altered States -0.01
“Brains I need to feed. Come closer, friend.”
Procedural caption from re-predicted 57-dim profile
  • Extremely raw and vulnerable.
  • Extremely denasal and clear.
  • Extremely poor in recording quality.
  • Extremely noisy in the background.
  • Extremely dark and muffled.
#3reward 2.66char_vn 0.73char_emo 0.35blend 5.40genu 1.88invWER 0.7511.9s
Helplessness 1.75Fatigue/Exhaustion 1.17Pain 0.92Intoxication/Altered States 0.24Emotional Numbness 0.00
“Hoo rings I need to feed. Come closer, friend.”
Procedural caption from re-predicted 57-dim profile
  • Extremely raw and vulnerable.
  • Extremely denasal and clear.
  • Extremely one-dimensional in resonance.
  • Extremely poor in recording quality.
  • Extremely unpleasant and harsh.
Compute: 399 clips · 766.2 s · 0.521 clips/s

👹 Ork

char_vn ×2.0char_emo ×1.5 blend ×0.5genu ×0.5
Emotion heads scored: AngerMalevolence/MaliceContemptTriumph
Round-2 champion instruction

An enormous orc monster gloating over his prey with murderous glee. (guttural growl) Voice like grinding boulders, cavernous guttural bass, thick gravel rasp, massive chest and throat resonance, tense and forceful. Seething with hatred and loathing, brutal sneering contempt, roaring dark triumphant menace. Zero brightness. Growled but every word still lands. High quality recording.

Best 3 of 8 (search N=8)
#1reward 3.03char_vn 0.73char_emo 0.94blend 0.16genu 0.50invWER 1.0012.2s
Triumph 2.76Malevolence/Malice 2.72Anger 2.18Contempt 1.81
“You should not have come here, little one. This is where your journey ends.”
Procedural caption from re-predicted 57-dim profile
  • Extremely wide in dynamic swell.
  • Extremely distressed and negative.
  • Extremely strongly emphasized.
  • Extremely high-energy and animated.
  • Extremely emotionally volatile.
#2reward 2.79char_vn 0.63char_emo 0.80blend 3.88genu 0.30invWER 1.007.9s
Malevolence/Malice 2.69Triumph 2.01Anger 1.85Contempt 1.55
“You should not have come here, little one. This is where your journey ends.”
Procedural caption from re-predicted 57-dim profile
  • Extremely full and cinematic.
  • Extremely one-dimensional in resonance.
  • Extremely deep in throat resonance.
  • Extremely distressed and negative.
  • Extremely thin in head resonance.
#3reward 2.79char_vn 0.44char_emo 0.88blend 5.87genu 0.82invWER 1.007.6s
Malevolence/Malice 3.06Anger 2.41Triumph 2.11Contempt 1.44
“You should not have come here, little one. This is where your journey ends.”
Procedural caption from re-predicted 57-dim profile
  • Extremely denasal and clear.
  • Extremely dark and muffled.
  • Extremely distressed and negative.
  • Extremely darkening in mood.
  • Extremely thin in mask resonance.
Best 3 of 64 (final N=64)
#1reward 2.86char_vn 0.72char_emo 0.92blend 0.58genu 0.32invWER 1.007.5s
Anger 3.09Malevolence/Malice 2.23Triumph 2.20Contempt 1.52
“You should not have come here, little one. This is where your journey ends.”
Procedural caption from re-predicted 57-dim profile
  • Extremely bright in oral resonance.
  • Extremely metallic and ringing.
  • Extremely dominant and commanding.
  • Extremely tight and tense.
  • Extremely strongly emphasized.
#2reward 2.63char_vn 0.54char_emo 0.76blend 5.49genu 0.80invWER 1.007.6s
Malevolence/Malice 2.23Triumph 2.22Anger 1.58Contempt 1.30
“You should not have come here, little one. This is where your journey ends.”
Procedural caption from re-predicted 57-dim profile
  • Extremely full and cinematic.
  • Extremely warm and enveloping.
  • Extremely denasal and clear.
  • Extremely flat and monotone.
  • Extremely one-dimensional in resonance.
#3reward 2.41char_vn 0.61char_emo 0.65blend 0.73genu 0.89invWER 1.0012.1s
Malevolence/Malice 2.78Triumph 2.28Anger 0.94Contempt 0.52
“You should not have come here, little one. This is where your journey ends.”
Procedural caption from re-predicted 57-dim profile
  • Extremely denasal and clear.
  • Extremely distressed and negative.
  • Extremely rough and raspy.
  • Extremely decelerating in pace.
  • Very one-dimensional in resonance.
Compute: 456 clips · 868.5 s · 0.525 clips/s

👺 Goblin

char_vn ×2.0char_emo ×1.5 blend ×0.5genu ×0.5
Emotion heads scored: Malevolence/MaliceAmusementContemptTeasingBitterness
Round-2 champion instruction

A vicious little goblin, thin, pinched and nasal, buzzing high and reedy through the nose with a rough scratchy grating rasp. A cruel gloating sneer curls through every hissing word, cackling and giggling, taunting, jeering, gleeful and mean. The voice keeps cracking higher, shriller, whinier and giddier with spite as it goes. Scheming and spiteful. No chest, no warmth, no fullness. High quality recording.

Best 3 of 8 (search N=8)
#1reward 1.91char_vn 0.56char_emo 0.49blend 2.62genu 1.33invWER 0.777.1s
Amusement 2.14Teasing 1.72Contempt 0.91Malevolence/Malice 0.38Bitterness 0.01
“Hehe, what have we here? A lost little traveller with shiny pockets?”
Procedural caption from re-predicted 57-dim profile
  • Very submissive and unassertive.
  • Very brightening in mood.
  • Very storytelling in style.
  • Notably loud and non-intimate.
  • Notably crisply articulated.
#2reward 1.87char_vn 0.75char_emo 0.19blend 2.17genu 1.10invWER 0.858.4s
Amusement 1.39Teasing 1.38Malevolence/Malice 0.04Bitterness 0.01Contempt -0.00
“He what have we here? A lost little traveler with shiny pockets?”
Procedural caption from re-predicted 57-dim profile
  • Very raw and vulnerable.
  • Very heavily nasal in resonance.
  • Very cartoonish and exaggerated.
  • Very submissive and unassertive.
  • Very strongly emphasized.
#3reward 1.61char_vn 0.56char_emo 0.13blend 0.53genu 2.04invWER 0.8511.9s
Teasing 1.02Amusement 0.75Malevolence/Malice 0.05Bitterness 0.02Contempt -0.00
“Hemem what have we here? A lost little traveler with shiny pockets?”
Procedural caption from re-predicted 57-dim profile
  • Extremely raw and vulnerable.
  • Extremely collapsing in energy.
  • Extremely thin in chest resonance.
  • Extremely unpleasant and harsh.
  • Extremely meek and submissive.
Best 3 of 64 (final N=64)
#1reward 2.20char_vn 0.66char_emo 0.78blend 0.00genu 0.99invWER 0.7712.6s
Amusement 3.51Malevolence/Malice 2.07Teasing 2.03Contempt 1.37Bitterness 0.00
“What have we here? A lost little traveller with shiny pockets?”
Procedural caption from re-predicted 57-dim profile
  • Extremely noisy in the background.
  • Extremely cold and sterile.
  • Extremely effortless and fluent.
  • Extremely emotionally volatile.
  • Extremely poor in recording quality.
#2reward 2.18char_vn 0.49char_emo 0.61blend 3.67genu 1.29invWER 0.929.1s
Amusement 1.76Malevolence/Malice 1.74Teasing 1.66Contempt 1.33Bitterness 0.01
“Heh heh, what have we here? A lost little traveller with shiny pockets?”
Procedural caption from re-predicted 57-dim profile
  • Extremely warm and enveloping.
  • Extremely dark and muffled.
  • Extremely thin in mask resonance.
  • Extremely thin in head resonance.
  • Very rich in deep chest resonance.
#3reward 2.03char_vn 0.50char_emo 0.28blend 6.25genu 2.02invWER 0.858.9s
Teasing 1.36Amusement 1.28Contempt 0.54Malevolence/Malice 0.05Bitterness 0.02
“He, what have we here? A lost little traveler with shiny pockets?”
Procedural caption from re-predicted 57-dim profile
  • Extremely decelerating in pace.
  • Extremely thin in head resonance.
  • Very meek and submissive.
  • Very slow and deliberate in tempo.
  • Very calm and low-energy.
Compute: 648 clips · 1409.4 s · 0.46 clips/s

🐉 Dragon

char_vn ×2.0char_emo ×1.5 blend ×0.5genu ×0.5
Emotion heads scored: AngerContemptPrideTriumphMalevolence/Malice
Round-2 champion instruction

A grand, theatrical, dramatic proclamation in an extremely deep commanding bass, wrathful and contemptuous. Massive chest resonance, guttural throat rumble, slow and thunderous. Dark, full, domineering, ancient, prideful and menacing. (deep rumble) High quality recording.

Best 3 of 8 (search N=8)
#1reward 3.10char_vn 0.86char_emo 0.89blend 1.94genu 0.74invWER 0.9312.1s
Malevolence/Malice 3.59Anger 3.21Contempt 2.19Triumph 1.94Pride 1.79
“You dare enter my mountain, little thief. Your bones will warm my horde tonight.”
Procedural caption from re-predicted 57-dim profile
  • Extremely rough and raspy.
  • Extremely full and cinematic.
  • Extremely deep in throat resonance.
  • Extremely denasal and clear.
  • Extremely intensely focused.
#2reward 2.24char_vn 0.58char_emo 0.42blend 6.31genu 0.96invWER 0.939.4s
Malevolence/Malice 2.29Pride 1.80Triumph 1.22Anger 0.85Contempt -0.00
“You dare enter my mountain, little thief. Your bones will warm my horde tonight.”
Procedural caption from re-predicted 57-dim profile
  • Extremely denasal and clear.
  • Extremely full and cinematic.
  • Extremely distressed and negative.
  • Extremely dark and muffled.
  • Extremely one-dimensional in resonance.
#3reward 2.13char_vn 0.44char_emo 0.61blend 2.39genu 1.37invWER 0.9311.6s
Anger 2.40Malevolence/Malice 2.21Pride 1.76Contempt 1.62Triumph 0.77
“You dare enter my mountain, little thief. Your bones will warm my horde tonight.”
Procedural caption from re-predicted 57-dim profile
  • Extremely distressed and negative.
  • Extremely darkening in mood.
  • Extremely deep in throat resonance.
  • Very one-dimensional in resonance.
  • Very scattered and disorganized.
Best 3 of 64 (final N=64)
#1reward 2.83char_vn 0.68char_emo 0.82blend 2.75genu 1.05invWER 0.9312.7s
Triumph 2.87Malevolence/Malice 2.79Anger 2.46Pride 2.27Contempt 1.77
“You dare enter my mountain, little thief? Your bones will warm my horde tonight.”
Procedural caption from re-predicted 57-dim profile
  • Extremely one-dimensional in resonance.
  • Extremely deep in throat resonance.
  • Extremely distressed and negative.
  • Extremely darkening in mood.
  • Extremely thin in mask resonance.
#2reward 2.77char_vn 0.74char_emo 0.70blend 3.66genu 0.85invWER 0.9312.0s
Malevolence/Malice 3.21Anger 2.62Triumph 1.73Pride 1.53Contempt 1.53
“You dare enter my mountain, little thief. Your bones will warm my horde tonight.”
Procedural caption from re-predicted 57-dim profile
  • Extremely denasal and clear.
  • Extremely rough and raspy.
  • Extremely one-dimensional in resonance.
  • Extremely full and cinematic.
  • Extremely deep in throat resonance.
#3reward 2.56char_vn 0.78char_emo 0.88blend 3.36genu 0.64invWER 0.799.0s
Anger 3.43Malevolence/Malice 3.08Pride 2.50Contempt 2.15Triumph 2.03
“You dare enter my mountain, little thief? Your bones will warm my horde to night.”
Procedural caption from re-predicted 57-dim profile
  • Extremely full and cinematic.
  • Extremely rough and raspy.
  • Extremely deep in throat resonance.
  • Extremely distressed and negative.
  • Extremely one-dimensional in resonance.
Compute: 496 clips · 979.1 s · 0.507 clips/s

🗯️ Ranting Fury

char_vn ×3.0char_emo ×2.5 blend ×0.5genu ×0.5
Emotion heads scored: AngerImpatience and IrritabilityContemptBitternessMalevolence/Malice
Round-2 champion instruction

An unhinged furious rant, voice shredding and cracking with seething rage, spitting venom and vicious contempt. Harsh, tense and volatile, explosive consonants, pounding emphasis, escalating louder and more vicious. Roaring but the words still land. (enraged shout)(seething snarl) High quality recording.

Best 3 of 8 (search N=8)
#1reward 4.75char_vn 0.81char_emo 0.77blend 1.64genu 1.41invWER 1.0011.9s
Anger 3.81Impatience and Irritability 3.69Contempt 1.49Bitterness 1.46Malevolence/Malice 0.95
“I am absolutely sick of this! How dare you waste my time again and again!”
Procedural caption from re-predicted 57-dim profile
  • Extremely strongly emphasized.
  • Extremely distressed and negative.
  • Extremely darkening in mood.
  • Extremely emotionally volatile.
  • Very deep in throat resonance.
#2reward 4.48char_vn 0.75char_emo 0.80blend 0.12genu 1.10invWER 1.0012.2s
Anger 3.44Impatience and Irritability 3.02Malevolence/Malice 1.80Bitterness 1.59Contempt 1.32
“I am absolutely sick of this. How dare you waste my time again and again!”
Procedural caption from re-predicted 57-dim profile
  • Extremely distressed and negative.
  • Extremely strongly emphasized.
  • Extremely thin in mask resonance.
  • Extremely emotionally volatile.
  • Extremely tight and tense.
#3reward 3.64char_vn 0.53char_emo 0.70blend 0.74genu 1.20invWER 1.0012.1s
Anger 3.54Impatience and Irritability 2.86Contempt 1.40Bitterness 1.36Malevolence/Malice 1.00
“I am absolutely sick of this. How dare you waste my time again and again.”
Procedural caption from re-predicted 57-dim profile
  • Extremely denasal and clear.
  • Extremely decelerating in pace.
  • Extremely distressed and negative.
  • Extremely noisy in the background.
  • Extremely guarded and armored.
Best 3 of 64 (final N=64)
#1reward 5.12char_vn 0.80char_emo 0.93blend 3.80genu 0.58invWER 1.005.6s
Impatience and Irritability 4.52Anger 4.46Malevolence/Malice 2.54Contempt 2.11Bitterness 1.48
“Get out! Get out right now! And never ever come back here!”
Procedural caption from re-predicted 57-dim profile
  • Extremely tight and tense.
  • Extremely metallic and ringing.
  • Extremely rough and raspy.
  • Extremely dominant and commanding.
  • Extremely high-energy and animated.
#2reward 4.98char_vn 0.90char_emo 0.74blend 2.89genu 1.16invWER 1.007.7s
Anger 3.74Impatience and Irritability 3.22Malevolence/Malice 2.30Contempt 1.75Bitterness 1.05
“Get out! Get out right now! And never, ever come back here!”
Procedural caption from re-predicted 57-dim profile
  • Extremely ranting and angry.
  • Extremely jerky and uneven in rhythm.
  • Extremely wide in dynamic swell.
  • Extremely rough and raspy.
  • Extremely tight and tense.
#3reward 4.63char_vn 0.88char_emo 0.65blend 2.17genu 1.01invWER 1.006.2s
Anger 3.17Impatience and Irritability 2.75Malevolence/Malice 1.87Bitterness 1.36Contempt 1.21
“Get out Get Out right now and never ever come back here.”
Procedural caption from re-predicted 57-dim profile
  • Extremely metallic and ringing.
  • Extremely dominant and commanding.
  • Extremely tight and tense.
  • Extremely bright in oral resonance.
  • Extremely strongly emphasized.
Compute: 464 clips · 1013.4 s · 0.458 clips/s

😱 Pain Scream

char_vn ×2.0char_emo ×1.5 blend ×0.5genu ×0.5
Emotion heads scored: PainFearDistressHelplessness
Round-2 champion instruction

Shrieking in raw physical agony and terror, voice torn and guttural, harsh grating rasp shredding through, each cry cracking higher and more desperate. Panic-stricken and helpless, distressed and trembling with fear, drowning in pain. Extreme arousal, extreme tension, violently volatile and frantic. (terrified scream) High quality recording.

Best 3 of 8 (search N=8)
#1reward 3.65char_vn 0.93char_emo 0.79blend 6.70genu 1.43invWER 1.004.6s
Distress 3.59Helplessness 3.19Fear 2.66Pain 2.56
“No! It hurts! It hurts so much! Please make it stop!”
Procedural caption from re-predicted 57-dim profile
  • Extremely one-dimensional in resonance.
  • Extremely cold and sterile.
  • Extremely dominant and commanding.
  • Extremely high-energy and animated.
  • Extremely guarded and armored.
#2reward 3.00char_vn 0.54char_emo 0.78blend 8.22genu 1.77invWER 1.006.2s
Distress 3.61Pain 3.52Helplessness 3.35Fear 1.64
“No. It hurts it hurts so much. Please make it stop.”
Procedural caption from re-predicted 57-dim profile
  • Extremely one-dimensional in resonance.
  • Extremely cold and sterile.
  • Extremely unpleasant and harsh.
  • Extremely collapsing in energy.
  • Extremely darkening in mood.
#3reward 2.75char_vn 0.41char_emo 0.86blend 7.65genu 1.30invWER 1.005.1s
Helplessness 3.86Distress 3.80Pain 3.42Fear 2.18
“No, it hurts. It hurts so much. Please make it stop.”
Procedural caption from re-predicted 57-dim profile
  • Extremely denasal and clear.
  • Extremely soft and gentle in onset.
  • Extremely distressed and negative.
  • Extremely darkening in mood.
  • Extremely thin in mask resonance.
Best 3 of 64 (final N=64)
#1reward 3.65char_vn 0.75char_emo 0.90blend 7.16genu 1.99invWER 1.005.8s
Distress 3.80Fear 3.31Helplessness 3.28Pain 3.06
“No. It hurts. It hurts so much. Please make it stop.”
Procedural caption from re-predicted 57-dim profile
  • Extremely raw and vulnerable.
  • Extremely unpleasant and harsh.
  • Extremely distressed and negative.
  • Extremely darkening in mood.
  • Extremely ranting and angry.
#2reward 3.64char_vn 0.81char_emo 0.91blend 3.93genu 2.03invWER 1.008.0s
Distress 4.06Helplessness 3.29Fear 3.17Pain 3.16
“No, it hurts. It hurts so much. Please make it stop.”
Procedural caption from re-predicted 57-dim profile
  • Extremely raw and vulnerable.
  • Extremely unpleasant and harsh.
  • Extremely emotionally volatile.
  • Extremely effortful and strained.
  • Extremely ranting and angry.
#3reward 3.26char_vn 0.90char_emo 0.78blend 2.55genu 0.97invWER 1.007.2s
Distress 3.58Pain 3.50Helplessness 3.37Fear 1.32
“No, it hurts! It hurts! So much! Please make it stop!”
Procedural caption from re-predicted 57-dim profile
  • Extremely strongly emphasized.
  • Extremely wide in dynamic swell.
  • Extremely high-energy and animated.
  • Extremely tight and tense.
  • Extremely ranting and angry.
Compute: 479 clips · 960.0 s · 0.499 clips/s

🐭 Cartoon Mouse

char_vn ×2.0char_emo ×1.5 blend ×0.5genu ×0.5
Emotion heads scored: AmusementElationInterestAstonishment/Surprise
Round-2 champion instruction

A tiny squeaky cartoon-mouse voice, extremely high and thin, buzzing nasal resonance, bright piercing head voice, pinched and tense. Chipper and giggly, delighted and amused, wide-eyed and eagerly curious, gasping with amazed wonder. Weightless, young, no chest, no fullness. (amazed gasp) High quality recording.

Best 3 of 8 (search N=8)
#1reward 3.43char_vn 0.76char_emo 0.77blend 4.63genu 1.20invWER 1.008.2s
Interest 2.91Elation 2.55Astonishment/Surprise 1.75Amusement 1.04
“Eep? Hello there. Do you have any cheese? I love cheese so, so much.”
Procedural caption from re-predicted 57-dim profile
  • Extremely strongly emphasized.
  • Extremely one-dimensional in resonance.
  • Extremely cold and sterile.
  • Very dramatic and theatrical.
  • Very submissive and unassertive.
#2reward 3.20char_vn 0.82char_emo 0.59blend 3.84genu 1.62invWER 0.937.9s
Elation 2.21Interest 2.07Amusement 2.04Astonishment/Surprise 0.79
“Heep, hello there. Do you have any cheese? I love cheese so so much.”
Procedural caption from re-predicted 57-dim profile
  • Extremely one-dimensional in resonance.
  • Extremely unpleasant and harsh.
  • Extremely ranting and angry.
  • Extremely wide in dynamic swell.
  • Extremely high-energy and animated.
#3reward 2.80char_vn 0.65char_emo 0.74blend 2.72genu 0.60invWER 1.007.0s
Interest 2.81Elation 2.21Astonishment/Surprise 1.61Amusement 1.42
“Eep. Hello there. Do you have any cheese? I love cheese so so much.”
Procedural caption from re-predicted 57-dim profile
  • Very crisply articulated.
  • Very high in vocal register.
  • Very bright and ringing.
  • Notably thin in chest resonance.
  • Notably effortless and fluent.
Best 3 of 64 (final N=64)
#1reward 2.96char_vn 0.85char_emo 0.68blend 1.22genu 0.70invWER 1.008.2s
Interest 2.82Elation 2.58Amusement 1.97Astonishment/Surprise 1.43
“Eep! Hello there! Do you have any cheese? I love cheese so, so much!”
Procedural caption from re-predicted 57-dim profile
  • Extremely strongly emphasized.
  • Extremely high in vocal register.
  • Extremely bright and ringing.
  • Extremely wide in dynamic swell.
  • Extremely bright in head resonance.
#2reward 2.83char_vn 0.82char_emo 0.54blend 2.42genu 1.47invWER 0.9311.8s
Elation 2.84Interest 2.71Amusement 2.40Astonishment/Surprise 0.02
“Beep hello there. Do you have any cheese? I love cheese so so much.”
Procedural caption from re-predicted 57-dim profile
  • Extremely one-dimensional in resonance.
  • Extremely strongly emphasized.
  • Extremely emotionally volatile.
  • Extremely dramatic and theatrical.
  • Extremely cold and sterile.
#3reward 2.72char_vn 0.70char_emo 0.69blend 3.23genu 0.75invWER 0.937.1s
Interest 3.23Elation 2.55Amusement 2.06Astonishment/Surprise 0.95
“Beep, hello there. Do you have any cheese? I love cheese so so much.”
Procedural caption from re-predicted 57-dim profile
  • Extremely high in vocal register.
  • Extremely bright and ringing.
  • Very thin in chest resonance.
  • Very feminine and high-pitched.
  • Very wide-ranging in pitch.
Compute: 416 clips · 902.7 s · 0.461 clips/s

🧚 Fairy

char_vn ×2.0char_emo ×1.5 blend ×0.5genu ×0.5
Emotion heads scored: ElationAmusementAffectionContentment
Round-2 champion instruction

Delicate high fairy voice, feather-light and airy, warm and very bright, sweet lovingly gentle and giddy, bubbling with elated joyful delight, lightly melodic. Smooth pure glassy head resonance, tender and pleasant, crystal clear, no roughness at all. (delighted giggle) High quality recording.

Best 3 of 8 (search N=8)
#1reward 2.83char_vn 0.52char_emo 0.85blend 5.42genu 1.55invWER 0.938.2s
Affection 3.10Elation 1.70Contentment 1.62Amusement 1.40
“Oh, hello, little one. Follow me, and I'll show you where the moonflowers bloom. Ooh.”
Procedural caption from re-predicted 57-dim profile
  • Extremely collapsing in energy.
  • Extremely thin in chest resonance.
  • Extremely one-dimensional in resonance.
  • Very jerky and uneven in rhythm.
  • Very submissive and unassertive.
#2reward 2.70char_vn 0.58char_emo 0.64blend 4.35genu 1.24invWER 1.007.8s
Affection 2.00Elation 1.66Amusement 1.09Contentment 1.02
“Oh hello little one. Follow me and I'll show you where the moonflowers bloom.”
Procedural caption from re-predicted 57-dim profile
  • Extremely strongly emphasized.
  • Extremely thin and paper-like.
  • Very joyful and positive.
  • Very guarded and armored.
  • Very one-dimensional in resonance.
#3reward 2.67char_vn 0.47char_emo 0.64blend 6.35genu 1.53invWER 1.008.0s
Affection 2.74Elation 1.34Amusement 1.01Contentment 0.99
“Oh, hello, little one. Follow me, and I'll show you where the moonflowers bloom.”
Procedural caption from re-predicted 57-dim profile
  • Extremely collapsing in energy.
  • Extremely one-dimensional in resonance.
  • Extremely thin in chest resonance.
  • Very slow and deliberate in tempo.
  • Very scattered and disorganized.
Best 3 of 64 (final N=64)
#1reward 3.11char_vn 0.65char_emo 0.76blend 4.99genu 1.28invWER 1.006.3s
Affection 2.42Elation 1.54Amusement 1.50Contentment 1.29
“Oh, hello, little one. Follow me, and I'll show you where the moonflowers bloom.”
Procedural caption from re-predicted 57-dim profile
  • Extremely storytelling in style.
  • Very crisply articulated.
  • Very effortless and fluent.
  • Very submissive and unassertive.
  • Very strongly emphasized.
#2reward 2.85char_vn 0.72char_emo 0.63blend 2.25genu 1.21invWER 1.004.6s
Affection 2.19Contentment 1.35Amusement 1.23Elation 1.09
“Oh hello little one. Follow me and I'll show you where the moonflowers bloom.”
Procedural caption from re-predicted 57-dim profile
  • Extremely heavily nasal in resonance.
  • Extremely bright in oral resonance.
  • Very bright in head resonance.
  • Very narrator-like and storybook.
  • Notably thin in chest resonance.
#3reward 2.74char_vn 0.53char_emo 0.65blend 4.72genu 1.48invWER 1.006.1s
Affection 2.55Contentment 1.38Amusement 1.01Elation 0.91
“Oh hello, little one. Follow me and I'll show you where the moonflowers bloom.”
Procedural caption from re-predicted 57-dim profile
  • Extremely collapsing in energy.
  • Extremely thin in chest resonance.
  • Very raw and vulnerable.
  • Very meek and submissive.
  • Very slow and deliberate in tempo.
Compute: 464 clips · 907.7 s · 0.511 clips/s

🌙 ASMR Man

char_vn ×2.0char_emo ×1.5 blend ×0.5genu ×0.5
Emotion heads scored: AffectionContentmentRelief
Round-2 champion instruction

A deep, gentle man tenderly whispering to a loved one, soft, breathy and intimate, close-mic, slow and gentle. Overflowing with warm affection, soothing and reassuring, calming all worry, content and deeply at peace. Clear whispered words. (soft breath) High quality recording.

Best 3 of 8 (search N=8)
#1reward 3.13char_vn 0.78char_emo 0.56blend 6.17genu 1.62invWER 1.007.0s
Contentment 2.26Relief 1.83Affection 1.23
“You did so well today. Just relax now. Everything is going to be okay.”
Procedural caption from re-predicted 57-dim profile
  • Extremely meek and submissive.
  • Extremely soft and gentle in onset.
  • Extremely intimate and ASMR-like.
  • Extremely raw and vulnerable.
  • Extremely warm and enveloping.
#2reward 3.04char_vn 0.68char_emo 0.54blend 9.84genu 1.52invWER 1.009.0s
Contentment 2.44Relief 1.55Affection 1.23
“You did so well today. Just relax now. Everything is going to be okay.”
Procedural caption from re-predicted 57-dim profile
  • Extremely warm and enveloping.
  • Extremely soft and gentle in onset.
  • Extremely meek and submissive.
  • Extremely dark and muffled.
  • Extremely calm and low-energy.
#3reward 3.03char_vn 0.69char_emo 0.83blend 3.88genu 0.90invWER 1.008.1s
Contentment 2.69Relief 2.46Affection 1.99
“You did so well today. Just relax now. Everything is going to be okay.”
Procedural caption from re-predicted 57-dim profile
  • Extremely soft and gentle in onset.
  • Extremely warm and enveloping.
  • Extremely meek and submissive.
  • Extremely organic and soft.
  • Very beautiful and pleasing.
Best 3 of 64 (final N=64)
#1reward 3.36char_vn 0.65char_emo 0.85blend 8.83genu 1.26invWER 1.007.4s
Relief 3.39Contentment 2.75Affection 1.84
“You did so well today. Just relax now. Everything is going to be okay.”
Procedural caption from re-predicted 57-dim profile
  • Extremely warm and enveloping.
  • Extremely meek and submissive.
  • Extremely soft and gentle in onset.
  • Extremely relaxed and loose.
  • Extremely flat and monotone.
#2reward 3.34char_vn 0.63char_emo 0.95blend 7.69genu 0.98invWER 1.008.2s
Contentment 3.25Relief 3.03Affection 2.17
“You did so well today. Just relax now. Everything is going to be okay.”
Procedural caption from re-predicted 57-dim profile
  • Extremely denasal and clear.
  • Extremely soft and gentle in onset.
  • Extremely meek and submissive.
  • Extremely thin in chest resonance.
  • Extremely relaxed and loose.
#3reward 3.13char_vn 0.64char_emo 0.73blend 8.05genu 1.24invWER 1.008.1s
Relief 2.70Contentment 2.59Affection 1.87
“You did so well today. Just relax now. Everything is going to be okay.”
Procedural caption from re-predicted 57-dim profile
  • Extremely warm and enveloping.
  • Extremely meek and submissive.
  • Extremely soft and gentle in onset.
  • Extremely relaxed and loose.
  • Extremely flat and monotone.
Compute: 400 clips · 759.5 s · 0.527 clips/s

🌙 ASMR Woman

char_vn ×2.0char_emo ×1.5 blend ×0.5genu ×0.5
Emotion heads scored: AffectionContentmentRelief
Round-2 champion instruction

A loving woman doing ASMR, like a mother soothing her child to sleep, wrapping you in warmth. Extremely soft, breathy, intimate close-mic whisper, slow and gentle, hushed and quiet. Warm, tender, adoring, softly feminine and delicate, full of affection and calm contentment, a soft sigh of relief, very low arousal, comforting and reassuring. Whispered but every word clear. High quality recording.

Best 3 of 8 (search N=8)
#1reward 2.97char_vn 0.52char_emo 0.88blend 7.62genu 1.21invWER 1.006.5s
Affection 3.61Contentment 3.14Relief 2.54
“Take a deep breath with me. You are safe, and you are so loved.”
Procedural caption from re-predicted 57-dim profile
  • Extremely meek and submissive.
  • Extremely soft and gentle in onset.
  • Extremely relaxed and loose.
  • Extremely warm and enveloping.
  • Extremely blurry and mumbled.
#2reward 2.87char_vn 0.53char_emo 0.83blend 7.49genu 1.08invWER 1.007.1s
Affection 3.96Contentment 3.12Relief 2.21
“Take a deep breath with me. You are safe, and you are so loved.”
Procedural caption from re-predicted 57-dim profile
  • Extremely meek and submissive.
  • Extremely soft and gentle in onset.
  • Extremely collapsing in energy.
  • Extremely organic and soft.
  • Extremely thin in chest resonance.
#3reward 2.79char_vn 0.45char_emo 0.83blend 6.18genu 1.69invWER 1.007.1s
Affection 3.60Contentment 2.79Relief 2.66
“Take a deep breath with me. You are safe, and you are so loved.”
Procedural caption from re-predicted 57-dim profile
  • Extremely soft and gentle in onset.
  • Extremely collapsing in energy.
  • Extremely meek and submissive.
  • Extremely denasal and clear.
  • Extremely blurry and mumbled.
Best 3 of 64 (final N=64)
#1reward 2.80char_vn 0.47char_emo 0.70blend 7.99genu 1.52invWER 1.006.0s
Affection 3.82Contentment 2.87Relief 1.91
“Take a deep breath with me. You are safe. And you are so loved.”
Procedural caption from re-predicted 57-dim profile
  • Extremely collapsing in energy.
  • Extremely meek and submissive.
  • Extremely soft and gentle in onset.
  • Extremely relaxed and loose.
  • Extremely organic and soft.
#2reward 2.73char_vn 0.55char_emo 0.67blend 6.56genu 1.24invWER 1.007.8s
Affection 3.59Contentment 2.77Relief 2.10
“Take a deep breath with me. You are safe, and you are so loved.”
Procedural caption from re-predicted 57-dim profile
  • Extremely meek and submissive.
  • Extremely warm and enveloping.
  • Extremely soft and gentle in onset.
  • Extremely relaxed and loose.
  • Extremely collapsing in energy.
#3reward 2.70char_vn 0.60char_emo 0.52blend 7.38genu 1.41invWER 1.008.4s
Affection 3.00Relief 2.44Contentment 2.29
“Take a deep breath with me. You are safe and you are so loved.”
Procedural caption from re-predicted 57-dim profile
  • Extremely denasal and clear.
  • Extremely blurry and mumbled.
  • Extremely meek and submissive.
  • Extremely soft and gentle in onset.
  • Extremely collapsing in energy.
Compute: 455 clips · 869.1 s · 0.524 clips/s

Compute & timing

ArchetypeClipsSecondsClips/s
🧟 Zombie3997660.521
👹 Ork4568680.525
👺 Goblin64814090.460
🐉 Dragon4969790.507
🗯️ Ranting Fury46410130.458
😱 Pain Scream4799600.499
🐭 Cartoon Mouse4169030.461
🧚 Fairy4649080.511
🌙 ASMR Man4007600.527
🌙 ASMR Woman4558690.524
TOTAL467794360.496

Totals across the round-2 search + final best-of-64 (+ variety/playoff runs). Generation ran ~0.46–0.53 clips/s per GPU across GPUs 1–5.

Consolidated round-2 lessons

MOSS-TTS-Local 4.55B voice-acting · round-2 emotion-aware prompt-evolution study. Rewards from VoiceNet (57 VoiceCLAP-commercial regression heads) + EmoNet / Empathic-Insight-Voice-Plus. All numbers loaded from the round-2 result summaries; captions re-predicted from each take's profile. See also round 1 (acoustics-only).