Многоголосая озвучка диалогов через Dia TTS

★ 7.7 · content

dialogue-audio is a Claude Code skill that generates realistic two-speaker audio using Dia TTS via the belt CLI, with ElevenLabs integration for post-production. Speakers are tagged [S1] and [S2], and emotional delivery is shaped by punctuation — commas for short breaths, ellipses for hesitation, exclamation marks for energy — plus parenthetical non-speech cues like (laughs), (sighs), and (gasps). Built-in prompt patterns cover interview, tutorial, and debate formats, making it practical for podcasts, audiobooks, explainer videos, and character voice acting. Pacing is controlled through sentence length and a defined pause hierarchy ranging from ~0.3 seconds for commas to ~1.0 second for ellipses.