Audio Theater

Generate multi-character audio (radio drama, lipsync voice reference, or podcast) from a script or idea. The engine is the same in all modes: one clean Gemini TTS clip per line, one voice per character. Those clean per-line clips are exactly what you need to mix a radio play, feed lipsync in seedance-2, or assemble a podcast.

Setup

Install dependencies (one-time):

pip3 install -r ~/.cursor/skills/audio-theater/scripts/requirements.txt

Keys are reused from sibling skills - you usually do not need to set anything:

• Gemini API key from the asset-generator skill (gemini_api_key).
• Replicate API token from the sound-effects skill (replicate_api_token, with the usual fallbacks).
• (recommended for realistic SFX) ElevenLabs API key (elevenlabs_api_key) — get a free key at https://elevenlabs.io and set it with setup_key.py --elevenlabs KEY. When present, the SFX backend defaults to ElevenLabs (much more realistic foley/ambience than Stable Audio). Without it, SFX fall back to the sound-effects skill (Stable Audio).

Import/verify them once: