JBI — Laugh Injection Showcase

Line: "You know the funny thing about this model? [laugh] It absolutely refuses to laugh on its own. So we taught it — with a real one. [laugh]"
Base TTS: laion/moss-tts-local-transformer-4.55b-voice-acting-v2 · Laughs: human-recorded, RVC-converted per voice (index_rate=0 on non-speech) · Orpheus arm: unsloth/orpheus-3b-0.1-ft inline tags, whole-utterance RVC.

Orpheus → RVC — one throat (speech+laughs generated as a single stream, then whole-file voice conversion)

showcase_orpheus_marcus_1.wav ⬇ download
showcase_orpheus_marcus_2.wav ⬇ download
showcase_orpheus_oliver_1.wav ⬇ download
showcase_orpheus_oliver_2.wav ⬇ download

MOSS v2 speech + injected real laughs (recorded via our collection app, RVC-converted, sequential assembly, spectral-matched)

showcase_marcus.wav ⬇ download
showcase_marcus_alt.wav ⬇ download
showcase_marcus_gap.wav ⬇ download
showcase_oliver.wav ⬇ download
showcase_oliver_alt.wav ⬇ download
showcase_oliver_gap.wav ⬇ download

Laugh cleanup A/B (raw converted | cleaned)

laughclean_ab_marcus.wav ⬇ download
laughclean_ab_oliver.wav ⬇ download