← Back to GPT Image 2 Prompts
Portraits/gpt-image-2
JSON Color Grading UGC Model
A workflow prompt that uses JSON color analysis from a reference photo to create consistent UGC-style product shots.
Prompt
Analyze this photo and give me a detailed JSON prompt that recreates it. Break down the color grading and every exact color in the photo. Then use that JSON as a reference to generate a person holding the product. Save the generated photo as a character reference and attach it to every future generation for facial consistency. The JSON controls the lighting and color grading while the model handles the character.
Why this works
- • JSON-structured prompts provide precise color and lighting specifications the model can follow.
- • Character reference chaining enables facial consistency across multiple product shots.
- • Separating color grading from character generation gives you independent control over each aspect.
Variables you can swap
- • reference photo
- • product type
- • color grading style
- • model pose
Practical tips
- • Use Claude Opus for the initial JSON analysis 鈥?it produces more detailed visual breakdowns.
- • The biggest tell of AI photos is flat colors and grain; this method addresses both issues.