Write AI prompts for film shots
Turn an approved shot plan into image and video requests. Keep the starting frame, action, references and model limits visible before generation.
Write AI prompts for film shots by deciding which single action a viewer should see. A two-second product shot may have room for a box opening, while lifting the watch and changing its strap would need another beat. A prompt that asks for all three can drift away from the approved edit.
The compile-generation-prompts workflow uses story decisions, assets and a shot plan to describe still frames and moving shots separately. Where a provider profile is available, the agent checks features such as reference images, duration and end frames. A provider-neutral brief can stay open until those limits are known.
What it does
- Describe one observable storyboard frame
- Define a movement from a starting to an ending state
- Give every reference image a specific role
- Check duration and framing against a model profile
- Flag an unsupported requirement for a production decision
Install and use
- Use a compatible agent, the original shared schemas and authorised source material. This planning stage requires an approved story and shot list. A media account is only needed later to execute requests.
- Ask, “Compile the box-opening shot. Keep the watch face and lighting consistent with this reference and flag any unavailable end-frame control.”
- Check the request against the neighbouring shots. Resolve missing capabilities before sending work to a paid generation service.
FAQ
Source
Made by zhangzhangco. Original skill: compile-generation-prompts, published on GitHub under the MIT licence.
View on GitHub