← ClaudeAtlas

talking-head-scriptinglisted

Write scripts for spokesperson and avatar video that fit the duration, sound like speech, and survive lip-sync generation. Use whenever the user is writing a script for a talking-head video, an avatar presenter, a voiceover, or has a script that runs too long or sounds written.
apimageorg/apimage-skills · ★ 0 · AI & Automation · score 72
Install: claude install-skill apimageorg/apimage-skills
# Scripting for Talking-Head Video Two constraints govern this, and both are unusual. **Cost is per second.** `generate_lip_sync` bills at 3 credits/second with a 30-second cap. A 30-second clip is 90 credits — a third of a Starter month. Every unnecessary word has a price. **The script has to sound spoken, not written.** Text that reads fine on a page sounds stilted from a mouth, and synthetic delivery amplifies that rather than hiding it. ## Word count by duration Natural speaking pace is roughly 2.3-2.6 words per second. Plan against the lower end — synthetic voices that rush sync worse. | Duration | Words | Credits | Fits | |---|---|---|---| | 8s | ~20 | 24 | One claim, or a hook | | 12s | ~30 | 36 | Hook plus one point | | 15s | ~38 | 45 | Problem, then solution | | 20s | ~50 | 60 | Three tight beats | | 30s | ~75 | 90 | A full short ad. The cap | **Write the script, read it aloud with a timer, then cut.** Not estimate — actually time it. Everyone overestimates what fits. ## Cut these first The words that always go, in order: 1. **Greetings.** "Hi everyone", "Hey guys" — two seconds of nothing 2. **Self-introduction.** Nobody asked yet 3. **Meta-narration.** "In this video I'm going to explain..." 4. **Brand name repetition.** Once is enough 5. **Qualifiers.** "Basically", "essentially", "actually", "really", "just" 6. **Setup for the point.** Start at the point 7. **Anything the visual already shows** A 30-second script cut of those is usually a 15-second scr