Generative Fill is quite poor

Question

Am i the only one that finds generative fill quite poor and not really useful?

I read all the tutorials, prompts, wording etc and generative fill still manages to get simple things wrong.

For example, I asked for "white silk", and it generates yellow, or red silk.

I asked for "passionfruit" and it generates this yellow blob

I asked for "mango slices", it generates a whole, distored rectangle mango.

How is it getting simple things wrong? Even when I try different selections, it still gets the shape wrong.

Is this generative fill going to get smarter or what?

Kevin Stohlmeyer · Answer

@Dante25551024f8ag @Trevor.Dennis not playing apologist here but also not being overly dramatic.

Original:

Mango Slices seems to give a pretty good representation of the fruit.

Sliced passion fruit (not one word) gave me several options - some were meh but others seem to be passable.

It will get better as the AI is trained up - Take into context that this was released less than 6 mo. ago.

Other AI options have been on the market much longer and also trained on larger (albeit unethical) sources.

Sign up

To post, reply, or follow discussions, please sign in with your Adobe ID.

Sign in to Adobe Community

To post, reply, or follow discussions, please sign in with your Adobe ID.

Scanning file for viruses.

This file cannot be downloaded