video-promptslisted
Install: claude install-skill popnow-ai/popnow-claude-plugin
# Video prompts
Write one prompt per shot. The goal is a clip that is usable on the first or second try, because every attempt is paid.
## Choose the model and settings first
Read `references/models.md` for each model's clip lengths, resolutions, aspect ratios and audio support. Pick the cheapest model and the lowest resolution that fit the shot, and say why:
- Drafts and iteration: a fast or mini tier at 480p or 720p.
- Final shots: the full model at 720p or 1080p. Use 4K or 2K only when the delivery really needs it.
- Long single takes (over 15 seconds): only models that support them, and only for slow, simple action.
- Shots with spoken lines or synced sound effects: a model with native audio, with audio turned on.
If the user has not chosen a model, recommend one per shot and offer a cheaper alternative.
## Prompt structure
Write in plain, concrete sentences, in this order:
1. **Shot and camera**: shot size, lens feel and movement. "Medium close-up, 50mm, slow dolly in."
2. **Subject**: paste the fixed description from the continuity sheet word for word.
3. **Action**: one main action with a clear start and end. "She lifts the cup, pauses, and smiles at the camera."
4. **Setting and light**: place, time of day, weather, light quality and direction.
5. **Style**: "photoreal, shot on 35mm film, soft grain" or "3D animated, soft clay look".
6. **Sound** (models with audio): ambience, effects, and any dialogue in quotes with who says it. Keep spoken lines short, about