Examples
Prompts for reference-based image generation
Use these examples as starting points, then adjust the subject, platform ratio, and review rules for your own GPT Image 2 Image Generator workflow.
Production prompt
Use the uploaded reference as the main composition, keep the product shape and color direction, improve lighting, create a polished ecommerce variation, 4:5.
Why it works
It names the job and protects stable subject identity, clear composition, and practical campaign ratios.
Variation prompt
Create three variations for reference-based image generation, keep the same subject and message, test different composition options, and preserve stable subject identity, clear composition, and practical campaign ratios.
Why it works
It asks for comparable directions instead of random alternatives.
Revision prompt
Use the best result as the base. Improve only lighting, crop, and empty text space. Do not change the main subject or commercial purpose.
Why it works
It isolates visible issues without restarting the workflow.
Publishing prompt
Prepare the final visual for web use, keep the target ratio, clear headline area, and review-safe details for creators, ecommerce teams, and marketers.
Why it works
It connects generation to the actual publishing context.