Strange watermarks/blurring on text with gpt-image-2 (medium or high)
Users are reporting significant issues with gpt-image-2 when generating images via API, specifically concerning "mushy" areas and blurring behind text.…
I wanted to regenerate an ad using gpt-image-2 and connected the API, since as far as I know - only the "low" setting is available in ChatGPT itself. However, with the API, I much more frequently get strange "mushy" areas or blurring behind the text; it looks very artificial and definitely isn't suitable for print.
What has your experience with the API been like? Are there any potential solutions?
It’s a shame when the design finally works on the fifth to seventh attempt getting it 80–90% right only for those stupid labeling and watermark issues to make it completely unusable.
Unfortunately, it involves many different sections of text not something that can be easily fixed on the fly in Affinity.