← sources

Reverse-engineering cinematic lighting: From reference image to Midjourney v6 & SDXL prompts (Looking for feedback)

source: reddit (reddit:StableDiffusion)published 2026-10-30 17:45ingested 2026-10-31 02:38

retained text

>Hi all,  
>  
>Whenever I see portraits with dual neon lighting or specific hair textures, I used to spend 20 minutes guessing the right prompt tokens.  
>  
>I created **Deconstruct AI** ([getdeconstruct.com](https://getdeconstruct.com)) to break down any image into lighting schemes, subject details, and optimized prompts across different models (MJ, SDXL, ComfyUI).  
>  
>Above is a real breakdown of an editorial portrait.  
>  
>If you have a minute, try it out on one of your favorite reference images (10 free test credits included) and let me know: **How close does the generated prompt get to the original mood?** Any critique is super helpful!

document fetched 2026-10-31 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1xjm5o1/reverseengineering_cinematic_lighting_from