LAION AI

1000 reinterpreted reference voices (cluster->SIDON->pitch->MOSS reinterpretation)

MOSS-VA-v2 conditioning manual: 40 emotions (best prompt + LoRA merge %, reward, correlations) — MD skill + HTML

LAION MOSS Local-Transformer 1.5 voice-acting model (4.55B, 48 kHz) — inference, fast-inference guide, best-of-64 audio demos

Reference-voice reinterpretation before/after preview

Voice reinterpretation training-data preview (original + top-10 expressive re-performances, procedural captions + scores)

±10% pitch-shift before/after grid for reference-voice de-identification

VoiceCLAP text->audio search over 50k annotated voice samples

Reference-voice search demo — 1336 Gemini-annotated character voices, semantic + BM25 search

Autonomous NSFW-emotion optimization grid (18+, synthetic)

Autonomous LLM-driven search agent for LoRA+prompt strategies on MOSS-VA-v2 voice acting TTS