Prompt Builder

Video and image models follow a prompt best when every instruction has a clear job. This builder keeps the parts separate: you write what happens in the shot, then choose the style, composition, camera motion, angle, lighting and lens. It assembles them in a fixed order and colors each part so you can see what the model is being told.

How to use it

1. Write the scene first

   Say who or what is on screen, where they are and what they do, in the order it happens. Two or three plain sentences work better than a list of adjectives.




2. Choose one camera instruction of each kind

   One shot size, one camera motion and one angle. Two competing motions in the same shot, such as a dolly in and a pan, usually produce neither.




3. Add light and lens

   Lighting changes the mood more than any other choice. The lens controls how much of the room you see and how flat or deep it looks.




4. Copy as text or JSON

   Use the text version for models that take one prompt box. Use the JSON version for models and pipelines that read structured prompts, or when you want to keep dialogue, audio and things to avoid in separate fields.

Text or JSON: which one?

Both versions carry the same information. A text prompt is one continuous block: style first, then the scene, then camera, light and lens. A JSON prompt puts each of those in its own named field, which makes long prompts easier to edit and reuse, and stops a dialogue line or a negative prompt from being read as part of the scene description.

If you are not sure which your model prefers, start with text. Move to JSON when a prompt grows past a paragraph or when you are generating several shots that share the same style and audio settings.

Related references

Not sure which option to pick? The shot types and camera angles library explains every framing in the builder, the cinematic looks and lighting library covers the lighting options, and the prompt structure guide walks through a full example.