voiceover-productionlisted

# Voiceover & Native Audio (generation-time authority) This skill decides how an AI video gets its audio + voice **as it is generated**. `ugc-product-video`, `generic-video`, `mimic-video`, and `stitching-multi-clip` load it BY NAME — it is **not a standalone entry point**. **Scope: native audio + voice CARRY only.** An AI video's audio comes from the generation itself (native audio) and stays consistent across clips via the `@Audio1` carry. This skill does **NOT** mute footage to layer on a different voice. **Deliberately re-voicing an existing video — clone the original or pick a new voice, lip-sync the talking parts, mute + add audio on the rest — is the `voice-replacement` skill**, a separate flow the USER triggers explicitly *after* the video exists. If a rule about generation-time audio, voice carry, or native-audio defaults lives elsewhere, that place points here. ## The decision tree (key on the SOURCE of the footage) 1. **AI-generated clips — native audio, ALWAYS.** Set `generate_audio = true` on every clip on a model with native audio (Seedance 2.0, Veo 3.1). The spoken voice is baked into the generation. **Muting an AI generation is a defect**, not a "clean" result — never `generate_audio = false` to "add the voice later." 2. **Multi-clip (> ~15s) that needs ONE consistent voice — carry it via `reference-to-video`. This is the STANDARD multi-clip voice path, not an experiment, and you MUST attempt it:** - Generate clip-1 on `image-to-video` (