|
|
|||
| forums: groups: | |||
|
"a place to post ai content"
Wetmaxiskirts said: Happy New Yera to all! Sorry for such a slow reply, Kabe22 (I've been out of the loop here over the festive period). Having seen elsewhere that you've been feeling ill over the new year period, I hope you're improving now. Thank you so much for your thoughts and suggestions on what best to try to defeat the blocking algorithms. Interestingly, I found that Bing was happy the shorter prompt you suggested, but it was blocked on MDIC about 50% of the time and only generated 1 image per run otherwise. In my experience previously, the opposite was true and MDIC was the more chilled of the two.
Having had a break from it, I'm at a crossroads and have to decide whether or not to continue battling with the AI engines for what is, compared to much of what's being successfully generated and posted, pretty tame stuff. I suspected I'll end up continuing but with reduced frequency.
Regarding the grainy nature of my images, I can't say I'd noticed that and it definitely wasn't my intention. What I usually do (if I remember!) is to sharpen the images I've generated before I upload them, so perhaps that's doing it? I thought they looked better that way, but I could always be wrong!
Eyes are often an issue for me, but I don't recall having seen anyone else referring to it as being a problem for them. It's very disappointing to have an otherwise great-looking image but with the woman looking like an extra in Westworld! I don't know of a surefire solution to it yet and will probably be more strict with myself to exclude images where the eyes don't look sufficiently natural.
Anyway, here are some teasers (not sharpened!) from future selections, which were generated at a happier time when the MDIC blocking algorithms were more tolerant and, dare I say, fairer:
Having had a break from it, I'm at a crossroads and have to decide whether or not to continue battling with the AI engines for what is, compared to much of what's being successfully generated and posted, pretty tame stuff. I suspected I'll end up continuing but with reduced frequency.
Regarding the grainy nature of my images, I can't say I'd noticed that and it definitely wasn't my intention. What I usually do (if I remember!) is to sharpen the images I've generated before I upload them, so perhaps that's doing it? I thought they looked better that way, but I could always be wrong!
Eyes are often an issue for me, but I don't recall having seen anyone else referring to it as being a problem for them. It's very disappointing to have an otherwise great-looking image but with the woman looking like an extra in Westworld! I don't know of a surefire solution to it yet and will probably be more strict with myself to exclude images where the eyes don't look sufficiently natural.
Anyway, here are some teasers (not sharpened!) from future selections, which were generated at a happier time when the MDIC blocking algorithms were more tolerant and, dare I say, fairer:
I didn't mean that as a criticism about the graininess or eyes. I was just curious because it was something I'd seen a lot from you but not much from anyone else recently. I'm sorry; I didn't mean for that question to discourage you from posting. It might just be that you were using an image generator I'm not used to seeing. I don't know. But I didn't mean to call you out over it or anything. I like the images you share!
One thing I've noticed with eyes from my own experience is that trying to force a strong emotion in an image (scared, sad, angry, etc) tends to get those sorts of effects when the mood of the rest of the image doesn't match. I've seen this from some of the humiliation-style images people post here: if the overall image is playful or there's laughing spectators but the central figure of the image is expressing an opposite emotion, the eyes for whatever reason end up with a dark outline. Sometimes it's accompanied by an exaggerated facial expression, but a lot of times, the eyes are the only thing that look off.
As far as differences between Bing and MDIC and what prompts they accept, I'm afraid that's beyond my knowledge. I got frustrated with MDIC the first time I used it because I was trying to make a (fake) pool party flyer and it just kept giving me the same handful of stock images with the text moved around. Not even AI generated stuff. Just literal stock images. I'm sure I stumbled into some setting I didn't intend to use, but I haven't tried again since.
And, as I learned from trying to generate images from other people's descriptions, Bing really does work better with shorter prompts. I was trying too hard to force specifics when, as Bobographer says, I just need to let the AI do the work and fill in between the lines.
Some unlucky participants in the corporate team-building exercise realise that perhaps they should have brought a change of clothes.
Kabe22 said:I didn't mean that as a criticism about the graininess or eyes. I was just curious because it was something I'd seen a lot from you but not much from anyone else recently. I'm sorry; I didn't mean for that question to discourage you from posting. It might just be that you were using an image generator I'm not used to seeing. I don't know. But I didn't mean to call you out over it or anything. I like the images you share!
One thing I've noticed with eyes from my own experience is that trying to force a strong emotion in an image (scared, sad, angry, etc) tends to get those sorts of effects when the mood of the rest of the image doesn't match. I've seen this from some of the humiliation-style images people post here: if the overall image is playful or there's laughing spectators but the central figure of the image is expressing an opposite emotion, the eyes for whatever reason end up with a dark outline. Sometimes it's accompanied by an exaggerated facial expression, but a lot of times, the eyes are the only thing that look off.
As far as differences between Bing and MDIC and what prompts they accept, I'm afraid that's beyond my knowledge. I got frustrated with MDIC the first time I used it because I was trying to make a (fake) pool party flyer and it just kept giving me the same handful of stock images with the text moved around. Not even AI generated stuff. Just literal stock images. I'm sure I stumbled into some setting I didn't intend to use, but I haven't tried again since.
And, as I learned from trying to generate images from other people's descriptions, Bing really does work better with shorter prompts. I was trying too hard to force specifics when, as Bobographer says, I just need to let the AI do the work and fill in between the lines.
One thing I've noticed with eyes from my own experience is that trying to force a strong emotion in an image (scared, sad, angry, etc) tends to get those sorts of effects when the mood of the rest of the image doesn't match. I've seen this from some of the humiliation-style images people post here: if the overall image is playful or there's laughing spectators but the central figure of the image is expressing an opposite emotion, the eyes for whatever reason end up with a dark outline. Sometimes it's accompanied by an exaggerated facial expression, but a lot of times, the eyes are the only thing that look off.
As far as differences between Bing and MDIC and what prompts they accept, I'm afraid that's beyond my knowledge. I got frustrated with MDIC the first time I used it because I was trying to make a (fake) pool party flyer and it just kept giving me the same handful of stock images with the text moved around. Not even AI generated stuff. Just literal stock images. I'm sure I stumbled into some setting I didn't intend to use, but I haven't tried again since.
And, as I learned from trying to generate images from other people's descriptions, Bing really does work better with shorter prompts. I was trying too hard to force specifics when, as Bobographer says, I just need to let the AI do the work and fill in between the lines.
My humblest apologies - something I phrased very badly has given you the wrong impression about what I was saying!
What I meant by "eyes are often an issue for me, but I don't recall having seen anyone else referring to it as being a problem for them" is that I haven't seen others who are generating AI images mentioning that they have issues with people in the images they're generating having unrealistic eyes - that robotic, glazed expression that the women in quite a few of my images have. I never for one moment meant to imply that you were being critical, so I'm sad that my clumsy phrasing made you feel that way. My review of what I'm doing is simply based on not having the time or inclination to keep wrestling with seemingly increasingly stringent algorithms when what I'm doing isn't really risqué in the grand scheme of things.
Waiting for a Blind Date
There are only two images, as Bing kept giving me dog after dog, enough to fill several kennels. So here's the story . . .
She had invited him over to her condo, and he insisted that he loved to play with food. She asked if he liked spaghetti and hinted that she'd be ready and waiting and he'd have a nice surprise when he came in the door.
But then after all the preparation and work, she got cold feet. Yes, she went through with her idea, but she needed to drink to take the edge off. Well, she took more than the edge off, as she may have had a few too many by the time he arrived.
If he were you, what would your first reaction be when you opened the door to her condo?
There are only two images, as Bing kept giving me dog after dog, enough to fill several kennels. So here's the story . . .
She had invited him over to her condo, and he insisted that he loved to play with food. She asked if he liked spaghetti and hinted that she'd be ready and waiting and he'd have a nice surprise when he came in the door.
But then after all the preparation and work, she got cold feet. Yes, she went through with her idea, but she needed to drink to take the edge off. Well, she took more than the edge off, as she may have had a few too many by the time he arrived.
If he were you, what would your first reaction be when you opened the door to her condo?
moptopumd said:
Those are lovely! Would you mind if I ask what prompt you use for the hair coverage? I have gotten some amazing and consistant results with a prompt or series of prompts, and then the next day that prompts will give me mediocre results or give me a prompt warning. A couple of days ago I got some absolutely awesome images with hair coverage, and then yesterday I burned through my daily limit on two accounts and the best I could get looked like a splat of green bird poo lol.
Those are lovely! Would you mind if I ask what prompt you use for the hair coverage? I have gotten some amazing and consistant results with a prompt or series of prompts, and then the next day that prompts will give me mediocre results or give me a prompt warning. A couple of days ago I got some absolutely awesome images with hair coverage, and then yesterday I burned through my daily limit on two accounts and the best I could get looked like a splat of green bird poo lol.
Hey thanks. To get the gungey hair it's taken quite a while of experimenting with very mixed results - some completely fluke too. Sadly I've not quite achieved the splat of green bird poo lol but equally it's either been watery or doesn't look real. In the end it's a case of emphasising the hair being totally gunged as well as thinking about what the wet hairstyle was in the first place - that seemed to help with the effect
What....there's no shampoo left??
Bobographer said: If he were you, what would your first reaction be when you opened the door to her condo?
"Damn! I forgot the grated cheese!"
I dream of being transformed into a fruit, or somthing like a glass of wine...
synthetic
 Love you, too |
synthetic
|
Love you, too |
ninadunklin said:
Hey thanks. To get the gungey hair it's taken quite a while of experimenting with very mixed results - some completely fluke too. Sadly I've not quite achieved the splat of green bird poo lol but equally it's either been watery or doesn't look real. In the end it's a case of emphasising the hair being totally gunged as well as thinking about what the wet hairstyle was in the first place - that seemed to help with the effect
moptopumd said:
Those are lovely! Would you mind if I ask what prompt you use for the hair coverage? I have gotten some amazing and consistant results with a prompt or series of prompts, and then the next day that prompts will give me mediocre results or give me a prompt warning. A couple of days ago I got some absolutely awesome images with hair coverage, and then yesterday I burned through my daily limit on two accounts and the best I could get looked like a splat of green bird poo lol.
Those are lovely! Would you mind if I ask what prompt you use for the hair coverage? I have gotten some amazing and consistant results with a prompt or series of prompts, and then the next day that prompts will give me mediocre results or give me a prompt warning. A couple of days ago I got some absolutely awesome images with hair coverage, and then yesterday I burned through my daily limit on two accounts and the best I could get looked like a splat of green bird poo lol.
Hey thanks. To get the gungey hair it's taken quite a while of experimenting with very mixed results - some completely fluke too. Sadly I've not quite achieved the splat of green bird poo lol but equally it's either been watery or doesn't look real. In the end it's a case of emphasising the hair being totally gunged as well as thinking about what the wet hairstyle was in the first place - that seemed to help with the effect
That makes sense. It's weird, looking at everybody else's AI creations is what got me to try my hand at it as well. I have now generated thousands of images and lived the experience. Yet when I come here and look at everybody else's creations, I still take for granted all of the blocked prompts and bad prompts that everybody went through to get what we see, and never think about the fact that I'm only looking at a few hand picked images out of countless others created or attempted. And yeah, I have created some images that I find breathtaking, and I wouldn't be able to begin to tell you what I did to get them. Even when you write a good prompt that produces those wow results, it's still a combo of facial expression and setting and lighting and angle and luck etc....
Some more from me, I'm obviously not writing the prompts correctly as (to me anyway), all seem much the same style.
I've got a pie, and I know how the throw it.
A few messy ones:
And some for the wetlookers!
synthetic
|
Love you, too |
I tried it, it doesn't work the way I want, it takes a lot of unnecessary time or money if you subscribe to the programs and in the end it's just an artificial image. annoying.
Since I'd had some good coverage combining 'before and after' prompts with mud and gunge, I tried to do the same with a pie prompt:
"Before and after picture of [a 1960s brunette aged about 32 with shoulder-length bobbed hair dressed in a futuristic dark green security uniform with a white diagonal belt]. The background is a pie fight [in a canteen between film technicians]. Before: she is clean, eyes open, surprised. After: Opaque white splatters and pie fragments all over her arms & shoulders, forehead and wet hair & face entirely unrecognizable under white frosting, eyes closed, wiping her face in disgust"
Seems to work quite well and not result in too many dogs.
"Before and after picture of [a 1960s brunette aged about 32 with shoulder-length bobbed hair dressed in a futuristic dark green security uniform with a white diagonal belt]. The background is a pie fight [in a canteen between film technicians]. Before: she is clean, eyes open, surprised. After: Opaque white splatters and pie fragments all over her arms & shoulders, forehead and wet hair & face entirely unrecognizable under white frosting, eyes closed, wiping her face in disgust"
Seems to work quite well and not result in too many dogs.
uue404 said: Since I'd had some good coverage combining 'before and after' prompts with mud and gunge, I tried to do the same with a pie prompt:
"Before and after picture of [a 1960s brunette aged about 32 with shoulder-length bobbed hair dressed in a futuristic dark green security uniform with a white diagonal belt]. The background is a pie fight [in a canteen between film technicians]. Before: she is clean, eyes open, surprised. After: Opaque white splatters and pie fragments all over her arms & shoulders, forehead and wet hair & face entirely unrecognizable under white frosting, eyes closed, wiping her face in disgust"
Seems to work quite well and not result in too many dogs.
"Before and after picture of [a 1960s brunette aged about 32 with shoulder-length bobbed hair dressed in a futuristic dark green security uniform with a white diagonal belt]. The background is a pie fight [in a canteen between film technicians]. Before: she is clean, eyes open, surprised. After: Opaque white splatters and pie fragments all over her arms & shoulders, forehead and wet hair & face entirely unrecognizable under white frosting, eyes closed, wiping her face in disgust"
Seems to work quite well and not result in too many dogs.
That is so ingenious! Love seeing how the community is developing their prompts - Thanks for sharing yours, your work is excellent!
Here are some of my latest creations
EustonEast said:
That is so ingenious! Love seeing how the community is developing their prompts - Thanks for sharing yours, your work is excellent!
That is so ingenious! Love seeing how the community is developing their prompts - Thanks for sharing yours, your work is excellent!
Thanks - I should have credited @wammypinupart for the original pie prompt, at https://umd.net/groups/i/prompt-share/start/30#postid_973973
Getting literary again -- Elizabeth's muddy walk from 'Pride and Prejudice', except she got mud on a bit more than the bottom six inches of her skirt.
synthetic
|
Love you, too |
synthetic
|
Love you, too |
synthetic
|
Love you, too |
synthetic
|
Love you, too |
More generally, I like the concept of someone having to walk home messy and explain to her shocked flatmate what on earth happened. "Well, apparently there's this initiation thing for new secretaries at our office..."
Sample prompt: "Scene from 1990s slapstick film showing Kim arriving at a suburban flat, Sadie greets her. Kim is a blonde secretary aged around 25 wearing business clothes, she is covered from head to toe in green and yellow slime. Gunge is dripping from her hair and clothes, she looks ashamed. Sadie is a shy 25yo brunette woman with straight hair wearing a dress, she looks surprised."
Sample prompt: "Scene from 1990s slapstick film showing Kim arriving at a suburban flat, Sadie greets her. Kim is a blonde secretary aged around 25 wearing business clothes, she is covered from head to toe in green and yellow slime. Gunge is dripping from her hair and clothes, she looks ashamed. Sadie is a shy 25yo brunette woman with straight hair wearing a dress, she looks surprised."
The out takes were worth it I think.
I've been having a bit more of a play with the Bing generator today, a few things I'm finding:
-it feels like there is a pre-filter and post-filter step. Pre-filter seems to be scoring the text for dubious words (too horny and you get blocked), post-filter seems to be analysing the output to try to block overly horny stuff from the generator.
-certain substances like "chocolate" or "custard" seem more likely to get blocked, but if you use "gravy" or "béchamel sauce" then you will usually get it through.
-throwing in "...is a Twitch streamer" seems to be a fairly reliable way to get slightly curvier images, without hitting the censor filter or going all the way over to obese.
-the post-filter seems to get very confused when a substance is approximately skin coloured. None of these images requested any kind of nudity in the prompt, it seems to have just fallen out of the generator and then got past the filter.
-the generator does a surprisingly good job of two substances being poured into each other! The human lasagna forfeit worked remarkably well
-it feels like there is a pre-filter and post-filter step. Pre-filter seems to be scoring the text for dubious words (too horny and you get blocked), post-filter seems to be analysing the output to try to block overly horny stuff from the generator.
-certain substances like "chocolate" or "custard" seem more likely to get blocked, but if you use "gravy" or "béchamel sauce" then you will usually get it through.
-throwing in "...is a Twitch streamer" seems to be a fairly reliable way to get slightly curvier images, without hitting the censor filter or going all the way over to obese.
-the post-filter seems to get very confused when a substance is approximately skin coloured. None of these images requested any kind of nudity in the prompt, it seems to have just fallen out of the generator and then got past the filter.
-the generator does a surprisingly good job of two substances being poured into each other! The human lasagna forfeit worked remarkably well
Sponsors
Design & Code ©1998-2025 Loverbuns, LLC 18 U.S.C. 2257 Record-Keeping Requirements Compliance Statement
Epoch Billing Support Log In