Feedback on GPT-Image-1 Predefined Image Sizes

API Feedback
,
1

First of all, huge appreciation to the OpenAI team — the image generation quality of GPT Image 1 is amazing and the progress is clearly visible. Just a few suggestions to help improve it further:

  1. Object Cropping – In long or rich prompts, some key elements get cut off due to fixed sizes.
  2. Missing Details – Complex prompts often lose important visual info in the output.
  3. Warm Tone Bias – There’s a tendency towards brownie/warm tones even when not prompted.
  4. Canvas Limitation – Predefined sizes feel restrictive; custom aspect ratios would really help.

We understand the amount of hard work behind this — these small tweaks can make a big difference in output quality. Thanks for your efforts!

1 Like