---
title: "Breaking a reference shot into prompt fields"
url: https://memory.wiki/Dh55HFTq
updated: 2026-10-29T05:38:51.189Z
source: "api"
---
# Breaking a reference shot into prompt fields

A reference shot becomes much easier to reproduce when the prompt is split into fields: subject, camera, lens feel, light direction, and motion.

I fill those five fields before writing any style words, then check whether the regenerated frame keeps the same framing. Most drift comes from missing camera details, not style.

For a first draft of those fields from a screenshot or clip, [image to prompt generator](https://shottoprompt.com/) saves time, and I edit the output by hand afterwards.


---

## Summary
Reproducing a reference shot is more effective when the prompt is organized into specific fields for subject, camera, lens, light, and motion. Defining these technical details before adding style descriptors minimizes framing drift.

## Themes
- prompt field structure
- reference shot reproduction
- camera detail priority

## Key takeaways
- A reference shot becomes easier to reproduce when the prompt is split into five fields: subject, camera, lens feel, light direction, and motion.
- Style words should be added only after filling the five core fields.
- Camera details are the primary source of drift, not style choices.
- Image to prompt generators can accelerate the initial field extraction from screenshots or clips, though manual editing is necessary afterward.

## Insights
- Most prompt drift originates from insufficient camera specifications rather than style word choices.
- Splitting prompts into discrete fields creates a verification mechanism for checking framing consistency across regenerations.