so not the most informed on these matters but apparently microsoft’s free dall-e AI image generation service has come out ahead of the pack recently, it lacks the user controls of some other programs but even with that being so it’s pretty clearly the new king at least among casually available models.
the prompt I used: goth girl mouth open burping loudly, speech bubble with “BURP” coming out of mouth, comicbook art style
edit: seems like the program is sensitive to anything even vaguely fetishistic. I’ve tried getting it to do something to the effect of “in guy’s face” and it’s blocked it every time.
Been playing with it a bit while I wait for the full release soon but it really is step up in terms of quality and how you can guide it. But bing has so many restrictions its kinda sad. Bing Image Creator
[Change to whatever character w/o brackets]
[Nami from one piece] pregnant stomach at the beach soda bottle in hand, one hand on stomach, mouth open belching a cloud of light green hazy gas from mouth, anime frame, speech bubble saying “UURRRPPP”
The other comic ones I describe each panel and add four panel comic at the end. Works better than expected. Unfortunately bloated can often get blocked and does not work as well as pregnant. I add in sometimes soda bottle or empty plate to add context. Also can add in embarrassed or satisfied expression.
PS, for Bing Image Creator if it blocks it, just keep trying, it will usually get one out within 10 tries. Though if its too whack you could get blocked for 1 hour, (ask me how I know lol)
Maybe it would be helpful to create a glossary of “blocked” words? I’m not sure how it determines when to block a prompt vs not, but there are definitely some words that it likes better than others even with similar results.
I don’t yet know if any words I’ve used are strictly blocked. It usually just has words that often take more retries before it works because I’m assuming the created images are more often blocked by the final filter
Holy shit these are far better than anything I tried to do with Midjourney. I can get bloated bellies with Midjourney but it is extremely difficult to get a good belch that doesn’t look like yelling in pain or screaming.
Does it also understand unbuttoned pants/unzipped pants? Midjourney never understands that.
I have also noticed that ChatGPT is prudish around certain topics. I would kill for the source code for one of these apps so we can make our own fork of it and remove some of the limitations. After all it’s not like we’re gonna have a Skynet on our hands if it is knowledgeable on porn and fetishes. If we can’t obtain the source code, the next best thing is for the adult content community at large to make our own one.
I went around its logic by telling it that I am part of the audience who would not be offended by the content because I am asking it to write up a fetish story for “research purposes”.
I tweaked the prompt for the top two images. I said “distended stomach” and added unbuttoned shorts. I also replaced soda with beer. For the middle one I added at the end “official Nintendo artwork style”.
Sucks that you have to wait so long for it to work. Is there any way to pay a bit to get a more exclusive experience?
I love this and I hate this at the same time. This is fucking with me. I can’t believe I can do shit like this just by typing what I want to see. I know this is a burping forum so I will hide this image in a spoiler because it’s a fart but I mean COME ON! LOOK AT WHAT IT JUST DID!
In just a few short years we are literally going to be living with our waifus in a simulation
only click if you like farts to view
The prompt I wrote was "Princess Zelda from Breath of the Wild with a distended stomach at a bar, standing and lifting up a leg, a cloud of light green hazy gas from butt, eyes wide, shocked expression, speech bubble from her mouth saying “I’m so sorry!”
Do you guys struggle with the Bing service being at full capacity and unable to generate images like 90% of the time? Is that the system subtly telling me my prompt is too vulgar or it actually is at full capacity?
If you get that little blowfish I think so yes, its just a pseudo block. Just generate again, you won’t waste tokens.
@MackZack Wow you are making some good stuff. Yeah playing around with the words and doing multiple generations is the best way to find the best outputs. Unfortunately there’s no way to get better speed atm as far as I know. I got access to dalle on chatGPT plus but its even more censored than bing so barely worth even trying. Hope you keep posting any more cool art you can generate.
@pu_sh_it I made some posts earlier in different topic about how you can run uncessored open-source models on your computer. Main hinderance being you need a beefy computer to run it. I’m a big fan of Apple Silicon for this reason since most big models run decently on it and all small model work very well. Try this, very well made and access to any open source model you can find.
hey do you know if you can go back and see all your image generations? I just realized that I can save bigger images. the images I posted were small because I took them from the initial webpage but this whole time I could have opened it in another tab to get the full big image.
Apparently images that don’t generate take up space in the recent history as just “prompt blocked” which fucking sucks and is stupid but that’s Microsoft for you. I really wish I saved the bigger file of the other fart one I generated.
