Forum: AI Generated Art


Subject: Looking for a good image-to-image AI tool

KaratKarnie opened this issue on Jul 09, 2023 ยท 5 posts


KaratKarnie posted Sun, 09 July 2023 at 12:24 PM

Hi, I'm trying to find an AI art generation app, website, or downloadable software that satisfies the following criteria.  (At the same time; that is, activating one of these features doesn't exclude another.)


*Image to image generation:  Takes as input an image, a prompt, and a slider for how much to stick to the prompt vs the original image.

*Can generate photorealistic images, as opposed to "anime," "cartoon," etc.

*Either does not have a NSFW filter, or the filter is not too sensitive. (More about this in a minute.)

*Has an effective price of not more than about 0.03 USD per image.

*At the risk of overstating the obvious, actually produces decent-quality results.


For a while I thought I had a winner with the website https://www.zmo.ai, but a week or two ago that tool appears to have undergone a massive increase in the sensitivity of its NSFW filter which makes it very frustrating to use and no longer worth the money I was paying for it, so I'm looking for something new.  If anybody knows of a tool that is comparable to the photorealistic, image-to-image part of ZMO and does not have a NSFW filter, that's basically exactly what I'm looking for.

The first criteria seems to be the most limiting one.  Even when I specifically search for "image-to-image" or similar phrases, the vast majority of AI image generation tools I find are only text-to-image, not able to take a reference image as input.  But I figure that what I'm looking for must be out there somewhere.

Regarding NSFW filters, I don't think that the stuff I try to create is THAT risque.  I'm not exactly trying to submit prompts like, if you will pardon the illustration, "big-t***ed ebony pornstar shoves a dil*o in her dripping pu**y."  More like, "[ethnicity] lingerie model laying on a couch with [some synonym for "large breasts"], hourglass figure, and lacey red bra."  ZMO used to be able to handle stuff like that, at least with some tweaking of the wording, about 90% of the time, until recently.  But ZMO's filter as of now, and the filters of most other tools I've come across, seem to reject most any use of words like "breasts," "busty," curvy," "lingerie," "bikini," "bra," "stockings," etc, which is too restrictive for me.  (Or, alternatively, the prompt is accepted, but the post-generation filter blocks the resulting image nearly every time.)  ZMO is the only tool I've come across so far that had a filter but one that (formerly) wasn't too sensitive; but I've come across more than one tool that doesn't have a filter at all, it just fails one of the other criteria.

Regarding the price, 3 cents per image might not sound like a lot, but that's still more than 10x more expensive than what ZMO offered.  The package I was using cost around $30 a month and allowed 500 image generations per day, which worked out to about 0.2 cents per image.  I feel like I might be willing to go with 50 images per day at $50 per month, which would work out to about 3.33 cents per image; but that's about the limit of what I'd be comfortable with, and the only other tool I've found that (mostly) satisfies the other criteria costs more than twice that.

On a final note, to skip ahead a bit, I have an inkling that what I'm looking for can be satisfied by some form or other of "Stable Diffusion," but so far I can't make sense of Stable Diffusion.  Most of my attempts to look it up have resulted in links to multiple zip files, seemingly assuming that the user already knows how to cobble it together; and while I might be theoretically capable of that, I am simply not familiar with what it seems to be assuming I already know - and I'm saying this as someone with a master's degree in robotics engineering, it's not as though I don't know how to code.  I'm not saying that I'm at my wits' end in trying to look it up; I just wanted to get this point out there as I think might save some time.  The only version of Stable Diffusion that I've managed to actually use so far is this site, which produces atrocious-looking results or else I'm just not using it properly; this is the reason I added the fifth criteria above.

I feel like something which satisfies all of these criteria MUST be out there somewhere, and I just haven't been able to find it.  I could believe it if there were several options that met every criteria except the price, but I've only found tool like that so far and it has some other issues that make it less than ideal, so I figure there must be more out there.  Can somebody please help me out?  Much appreciated, thank you. :)


(Alternatively, as this is primarily a website about a 3D modeling software, perhaps somebody can suggest a better forum to be asking this question in.  I'm only asking it here because I've found surprisingly sparse results for "AI art forums.")