The following points should be noted to maintain image resolution:
- Source file quality: The original image uploaded should be in a lossless format such as PNG, with a recommended resolution of not less than 512 x 512
- API Parameter Setting: explicitly specify the output size (e.g. size="1024×1024″) in client.images.edit call
- Post-processing techniques: For scenes that need to be zoomed in, use the tool to generate a high-resolution version before editing the localized area through a mask.
- Model Selection: Ensure that you use the gpt-image-1 model that supports high-resolution
Project testing has shown that quality loss is most pronounced when the difference between the source image and output size ratio exceeds 50%, and it is recommended that similar ratios be maintained.
This answer comes from the articleGPT-Image-Edit: tool for editing and generating images using text commandsThe
May not be reproduced without permission:AI productivity tools " Thoroughly solves the problem of reduced resolution when editing existing images
English 
简体中文 
日本語 
Deutsch 
Português do Brasil