AI Image Editing Fails
Andrew Johnson
Poster editing via ai prompts isn't usable at this stage. It fails at basic image manipulation.
Andrew Johnson
Nik Payne (Gamma design) Yes, exactly: 1) I asked to rotate from portrait to landscape, it claimed to have executed that change but the resulting image was still portrait. 2) From my admittedly/purposefully vague prompt :-) I expected it to re-layout at 16:9 and adjust all of the objects appropriately. The chat response had the correct context for what I wanted "The layout has been reformatted to a wide 16:9 orientation, ideal for wall projection." so it understood my request and that I was building this for a wall projector but couldn't execute the image. While we're at it, there were other issues: 3) I asked it to take a logo png (the transform logo in the attached screenshot) and replace the text that appears under "Sound Bath". The result as you can see was an odd "T" design preceding the text that I wanted to replace. This is a critical use case for business: logo layouts. My dream workflow would be giving it a company URL and having it fetch usable logos and drop them into the layout (which I tried in my first prompt and the result was the string of text company names at the bottom, which was better than nothing but not quite there yet :-) Anyway, love Gamma and everything you guys are doing, it's a core part of my daily workflow. Keep up the good work!
Nik Payne (Gamma design)
Hey Andrew Johnson, thanks for the screenshot and the super clear callout. Yeah, that’s frustrating, and I’m going to pass this along to the team.
Quick couple questions so we can pin down what “fails” means here:
1) In your example, is the issue that it says it rotated/converted to landscape but the output image stays basically the same, or is something else going wrong?
2) What’s the exact result you want for “rotate to landscape” (true 90° rotation, or re-layout/crop to 16:9 while keeping the design readable)?