- This topic has 45 replies, 5 voices, and was last updated 2 weeks, 4 days ago by
.
-
Posts
-
-
By now, most of us have probably tried out one or more of the AI image generators. Some are terrible and a few are excellent. The feature they almost all share is a ban on nudity or anything remotely “adult” in nature. I’ve found one which is happy with nudity, produces high-quality images and also allows up to 100 free image generations per week. Obviously, there are limits (nothing illegal for starters) but if you type “Four naked women on the beach” then it generates exactly that (although there might sometimes be five!)There are paid features but 100 free images isn’t bad.
Diana, Ballsnwood and ptHarry like this
-
It’s a wild site. I’m unaccustomed to AI interfaces. My original and AI product are vastly and hilariously different.
Martin likes this
-
As Ed pointed out in another post (which I don’t seem to be able to link to) , the results are often impressive but contain obvious errors.
Often the results look correct when they are first seen:
but it only takes second glance to spot the mistakes:
- is that table floating in the pool?
- they appear to be playing a game, but what is it?
- why is the woman in the middle wearing a bright yellow glove on her right hand, apparently to take off her matching yellow bikini
- the man left of centre is wearing a brown shirt that is shaped like a lifejacket
Oh, I think I’ve worked it out. AI thinks the game is pool, So they’re playing it on what it thinks is a pool table, which is why it’s floating. And the man is wearing a life jacket because he’s worried it is going to sink 🙂
NakedMadCat and Martin like this
-
Ed’s post is at Ed’s post
But when I tried to update my previous message to include it, the whole post disappeared twice.
-
-
But it seems not to understand that cats don’t wear clothes.
To make it appropriate for the intended audience here, I changed the prompt to a naked black cat in a coal cellar at night:
Ed likes this
-
That is one weird interpretation. I think the large language models and latent diffusion models are an impressive leap forwards but they are NOT the general intelligence we’re striving for. People see more intelligence there than there actually is.
ptHarry likes this
-
The forum seems to have lost my Naked Black Cat in a Coal Cellar at Night picture that was originally in the message a few posts back,
I thought that might happen, because the picture was uploaded in the post that disappeared. But the upload was still on the server at the time. Looks like scheduled maintenance has removed it, so here it is again:
-
-
-
-
Other times it can be rather stupid:
Why would an artist mount her canvas on the wrong side of her easel?
Ed likes this
-
-
-
-
-
-
Onlyfakes has weekly competitions. This week, not too surprisingly, is about Valentines. Unusually, they have said this week’s entries should be Safe For Work.
So, vaguely inspired by Shakespeare in Love:
-
Or: (not entered for competition due to this week’s rules):
Act 2: The same, three hours later
Ed likes this
-
-
My Shakespeare in Love had 9 votes. More than I expected.
The winning entry was a teddy bear with a box of chocolates, getting 12,
Ed likes this
-
-
This was inspired by Knocking a door naked is much more daring than answering it!
(but it doesn’t seem to be allowed to post a link and an image in the same post, so see the next one).
-
-
-
-
This on is actually pretty hilarious. Original is me and my guy friend in a public pool. The filter turned him into my paw
ptHarry likes this
-
@flflirt: The thumbnail only shows part of the full size display:
Maybe the filter is trying to give him a message: “hands off” 🙂
-
It’s the details that almost always give AI-generated images away. This pic is almost perfect apart from the wispy beard. Where did that come from?
-
-
-
-
Ed: “don’t look at the guy on the right’s hands”
And the women behind him seems to be working for a company that only employs a skeleton staff 🙁
– – –
Sometimes, it works better. I uploaded a photo from the Albert Memorial:
and asked it to add Victorian costumes in full colour.
-
-
-
If we can dress them, then we ought to be able to undress them too.
But one of the men was too shy and colour photography was too expensive in 1860. , The Statues Union said the women needed body doubles and even then, one of them visibly turned their nose up about having to pose on a plinth in a public park.
-
But a Victorian artist would not have been restricted to painting them in black and white:
Ed likes this
-
-
-
-
-
-
Is anyone still experimenting with this?
-
Anything in particular you think we should try?
-
I did this one for a competition:
“Her birthday is April 1, so the new lion tamer at the local zoo has to wear her birthday suit and invite the lions for a picnic”.
Not recommended to do at home, but if she was real, she would qualify for the Naked at Work page 🙂
Ed likes this
-
-
-
That would explain it. I’m still using GetImg. I still can’t get good results out of it like I could before so it could well be them.
This is the best I can get now:
-
-
It’s not perfect but the results look nice:
There are still frequent problems with fingers and it’s often best not to look too closely at the background!
ptHarry likes this
-
-
-
Ed: “There are still frequent problems with fingers”
The story I heard was that AI gets confused because there are too many examples of wrongly-drawn hands on the web. AI can only “learn by example”. If it sees a significant proportion of incorrectly drawn hands on the web, and has no way to be taught right from wrong. And then, we can expect AI to generate a similar percentage of incorrectly drawn hands. Effectively, “garbage in, garbage out”.
Many of us have been looking at web pages for 25 years. Most of us have probably never remember seeing a web page and exclaiming “that hand is wrong”. Yet as soon as we judge created by AI, we notice. Could it be that we subconsciously assume that AI should be superior, and apply a higher standard to the work it produces than to the work produced by mere humans?
Maybe even, if I popped into the National Gallery and had a close look at paintings I’ve briefly glanced at before and study them in more detail, I would find similar examples of badly drawn hands amongst them. But I don’t have time to do that right at the moment.
Instead I did a web search, and found this:
Why can’t AI draw realistic human hands?
Eray Eliaçık
January 19, 2023One observation it makes is that “most cartoon hands only feature three digits and a thumb. Since we’re used to it, our brains don’t even register it”.
I never knew that, and was prepared to believe it to be true — but wanted convincing. The cartoons I saw at an early age would mostly have been Beano and Dandy. But an image search for Beano and Dandy Cartoons completely disproved that idea. Every hand I checked had the correct number of fingers :). Perhaps Beano and Dandy were two examples of cartoons that were “different to most” but it seems too much of a coincidence. Perhaps the statement is correct for the USA market and UK cartoonists are more strict on accuracy.
What about the backgrounds?
Two possible reasons. Sites have to share resources between everybody that is using them, so putting less computation power into generating good backgrounds means possibly more attention can be given to getting the main person right. Certainly it seems that asking for just one person is more likely to give good results than asking for several. And one user has recommended specifying “blank background” to improve rendering of the foreground.</div>
But I also wonder whether that’s a technique that AI might have learned from famous artists in museum pictures found on the web. There you will often find huge pictures with a few people in the foreground painted in very high detail, and lots more people in the background each painted with just a few brushstrokes to give an impression of a lot of people with very little detail each
Ed likes this
-
I’ve read (no idea where) that AI’s struggle to “understand” (not the right word but I don’t know the actual term) the mechanics of human fingers. There are so many pictures with hands in completely differing positions and poses that unless you already know there should be four fingers and one thumb, all those photos where less are visible (often due to angle) mislead the learning.
Personally, we’re a LONG way from a human-like intelligence. With picture and text generation, we’ve got something very easy to anthropomorphise and see far more intelligence than is truly present.
-
-
- You must be logged in to reply to this topic.