I'm trying to make images that look actually wet, with light reflecting off of the people as though they've just popped out of a pool of water.

Bing was good at it, but too many filters. I'm looking for a quick and easy solution that will make wet looking images.

Geminim2.5 is amazingly realistic in the representation of wet clothes even with light incidence

I used to use bing as well but then I discovered google whisk which is far superior and a lot less filters and also no character limit to describe the scene so much easier to do multiple people as well. Just a few examples of some of my creations

Gemini used to be great before Google decided to switch to their much worse "Nano Banana" or "Gemini 2.5 Flash" model, unfortunately with strict filters, less realism (still better than Bing however), low resolution and square pictures only - just like in the very beginnings of AI picture generation.

You could still try Google AI studio, where you still can select Imagen 4 for creation instead of the much worse However, unfortunately it also has much more strict filtering than Gemini with the older model used to have.

I just noticed the poor resolution, just in time!

That's why I also cancelled my subscription, along with all the other downsides I mentioned.

You may also use Whisk or flow with your AI Pro subscription, but only with a very limited amount of credits compared to the virtual unlimited (image) creations that were possible with Gemini. Too bad!

Hopefully they'll bring back Imagen for creation some day. The image button still says "create images with Imagen" but unfortunately that's a lie. They seem to have been switching back and forth for a wile, my second (free) account was able to use the better model while the paid one didn't, then for some time it worked again on both. But now they seem to have to restricted both to the new model.

@Bobographer: I hope you don't get this wrong. Gemini is not completely bad, it just was better before and that's frustrating. It can, however, still do some pretty nice wetlook, depending on what you like to see and if you don't mind the lower resolution. Here are some examples with the recent model which I think turned out very nice:

You can also try to enhance the images with an AI upscaler like upscayl (upscayl.org), which usually works well, but sometimes not, as can be seen here with the woman and her dress

Actually, Gemini still is not THAT bad. It's just worse than it has been before with the latest model. By reworking some of my older prompts (from when it used an older model and even stricter filters than now) I was able to get some really nice results, leaving the low resolution and square images on the downside. And it can't do wet jeans very good, which is too bad since I love that.

Gemini is also still very good in doing partial wetness, something many other AIs tend to struggle with.

Some recent examples:

Google Imogen image creator does some spectacular work if you coax it properly. Don't use the new model NanoBanana use the older flash image gen. Also there is a setting that will relax the safety requirement... something like "allow more" versus "allow less". Also use all ages setting so you do not get any 'kids'. (that sounds backwards but if you use adult only then it seems to jam you up more).

Where do you chose the older model? I only have the choice between 2.5 Flash and 2.5 Pro, and both use Nano Banana.

But I get the impression they've loosened the filters a bit and image quality also seems to improve slightly. It still can't do revealing and seethrough pictures like it could with Imagen however, and the resolution is still low and does square images only (or I did not yet find out how to change that, when editing uploaded images it usually keeps the original aspect ratio).

But maybe they're working on that also.

Here, however, are some sample of what I made with Imagen 4 in Gemini before. You may guess I preferred that to Nano Banana model.

I get varied results from the different generators I use. Some are good some days, then bad the next, often with virtualy the same prompts. Some days one site won't allow a prompt, the next day welcomes it.

Yeah, that's what I also experience from the commercial AI engines. Gemini however has been delivering very constant results for a longer time before they introduced the Nano Banana model.