AI Video Voice Design
This stage is required even when the voice is unresolved. Follow:
unresolved → provider_selected → voice_candidates_generated → preview_generated → user_approved → full_voiceover_generated → production_ready
Select a voice appropriate for a knowledgeable young professional: clear, calm, credible, energetic without sounding promotional. Generate a short representative preview and obtain user approval before rendering the full narration.
Support preset voices, user-supplied audio, or authorized voice cloning. For MiMo, follow references/mimo.md. Never commit, print or embed API keys. Never clone a voice without the user's ownership or explicit authorization.
Produce a voice brief, candidate notes, preview files, approval evidence, the full voiceover, and timing metadata. If the voice or timing changes, require downstream caption timing, audio integration, motion/audio and calibration to run again.