Ghost Voice Lab
← dreamA/B style test for Gemini TTS. Edit the style instructions, generate each variant, compare them, and vote. Votes accumulate per scene across sessions. Wear headphones.
“The resonance here is ancient. Let yourself be absorbed by it.”
A
no audio yet
B
no audio yet
Which style better captures the Ghost character for this scene?
Try swapping the default styles between A and B to isolate what each parameter does.
Useful contrasts: slow/formal vs. breathy/intimate · flat affect vs. expressive · direct vs. distant.
Votes are stored per-scene in localStorage — they persist across sessions.
Gemini TTS · fal-ai/gemini-tts · FAL_KEY · ~$0.01 per generation pairdesign notes: src/app/dream/59-gemini-voice-lab/README.md