Trying out this AI thing

[Theo]

i bet i'm not the only one hoping you're going to experiment with O Girl in a swamp

I'm down for that

[cerberus]

i bet i'm not the only one hoping you're going to experiment with O Girl in a swamp

Well yes, but this sent my mind off in a completely different direction and highlighted a massive opportunity missed, seemingly, by two studios. How about Kendra James as Red Sonja! Have we missed out, has the opportunity gone or is there still hope? We can but dream ...

... and, maybe play with AI:

Attachment removed.

Apologies for posting an image here, I don't usually post images on other people's threads unless invited, but it seemed the right place!

Oh, and I liked the latest Vera picture, nicely done.

[Duncan Edwards]

i bet i'm not the only one hoping you're going to experiment with O Girl in a swamp

I'm down for that

Love the idea. One thing to remember about tinkering with "the real thing" with models like Kendra is we already have her. I've played around with existing real work a lot because well, I've got a lot of it. Still doesn't seem quite right to try and improve what we already thought was good enough. I'm sort of stretching the point to make one but can you imagine DaVinci Photoshopping the Mona Lisa? Just doesn't seem right. Maybe later.

[Theo]

i bet i'm not the only one hoping you're going to experiment with O Girl in a swamp

Well yes, but this sent my mind off in a completely different direction and highlighted a massive opportunity missed, seemingly, by two studios. How about Kendra James as Red Sonja! Have we missed out, has the opportunity gone or is there still hope? We can but dream ...

... and, maybe play with AI:

Apologies for posting an image here, I don't usually post images on other people's threads unless invited, but it seemed the right place!

Oh, and I liked the latest Vera picture, nicely done.

Like the Kendra pic I tried doing her a while back, but Perchance apparently didn't know her very well

[cerberus]

i bet i'm not the only one hoping you're going to experiment with O Girl in a swamp

Well yes, but this sent my mind off in a completely different direction and highlighted a massive opportunity missed, seemingly, by two studios. How about Kendra James as Red Sonja! Have we missed out, has the opportunity gone or is there still hope? We can but dream ...

... and, maybe play with AI:

Apologies for posting an image here, I don't usually post images on other people's threads unless invited, but it seemed the right place!

Oh, and I liked the latest Vera picture, nicely done.

Like the Kendra pic I tried doing her a while back, but Perchance apparently didn't know her very well

Thanks, SD doesn't know her either, I had a to use ReActor, which uses a supplied pic of the required face and puts that in the generated images. It leaves the face shape as generated so it is a bit hit and miss, but sometimes it works well.

[cerberus]

i bet i'm not the only one hoping you're going to experiment with O Girl in a swamp

I'm down for that

Love the idea. One thing to remember about tinkering with "the real thing" with models like Kendra is we already have her. I've played around with existing real work a lot because well, I've got a lot of it. Still doesn't seem quite right to try and improve what we already thought was good enough. I'm sort of stretching the point to make one but can you imagine DaVinci Photoshopping the Mona Lisa? Just doesn't seem right. Maybe later.

I'm guessing this was directed at me, so I'll reply.

Glad you like the idea, I suspect it isn't likely now, but it would have been cool. I understand what you mean by '... tinkering with the "real thing" ...' , I thought long and hard about posting that image here and am in two minds as to whether I should take it down again. I would say that none of the material from either MPV etc and Jim's sites were used here, the facial features came from a publicity shot regarding her book and the rest is purely AI. It was in no way intended to "improve" what has already been done, it is basically a "what if" with an example image. I do understand the point you are trying to make, it is a concern of mine. At the moment most of the AI images posted can be identified as fake without too much difficulty, but AI is getting better and may soon be real enough to pass a detailed examination. As long as the stuff we do stays on this site there shouldn't be a problem, but I can't see that lasting and at some point some of our images are going to escape into the wild. If someone then treats one of our fake images as genuine and uses it for malicious purposes; well, the consequences could be serious.

I'm going to have to disagree with you on Da Vinci, he wasn't just a painter he was also an engineer and inventor, my guess would be that he would have been all over the technology available today and using it to the utmost to achieve the best possible results.

[Jim_Weathers]

Here is the bottom line to all of this. This whole AI thing is so foreign to me. I have zero idea how to take a seed image and put that seed image into another image. I have tons of photos that I can use as reference, but whenever I try to do that, I just get a bad AI image of the image. I don’t know how to take an image of say kendra, and put her into a swamp.

It is basically my stupidity in all of this. I am using mage.space, and I can seed an image, but it doesn’t seem to do what I want. So yeah, it would be great to be able to put O-girl, or a Yvonne Craig Batgirl into a quicksand situation, but I’m at a loss as to how to go about it. Just like anything else, this is another example of a new technology that I just don’t understand at all.

Any help from any of you out there would be helpful

[cerberus]

Here is the bottom line to all of this. This whole AI thing is so foreign to me. I have zero idea how to take a seed image and put that seed image into another image. I have tons of photos that I can use as reference, but whenever I try to do that, I just get a bad AI image of the image. I don’t know how to take an image of say kendra, and put her into a swamp.

It is basically my stupidity in all of this. I am using mage.space, and I can seed an image, but it doesn’t seem to do what I want. So yeah, it would be great to be able to put O-girl, or a Yvonne Craig Batgirl into a quicksand situation, but I’m at a loss as to how to go about it. Just like anything else, this is another example of a new technology that I just don’t understand at all.

Any help from any of you out there would be helpful

I'll have a go, I'm not using Mage.Space but I use Stable Diffusion (which Mage.Space uses as it's image generator) so everything I mention should be available in Mage.Space at some subscription level (probably Pro) but it may be named differently.

There are several ways to approach this, I'll run through a couple. Note: results my not be exactly as you would like but it should be possible to get pretty close.

First method: This is using the Txt2Img option with ControlNet and ReActor.

Firstly you need a good prompt/negative prompt that gets you the sort of image you are looking for.

You could try this one and see how it works for you:

(horrified:0.9) attractive auburn woman wearing purple (heavy neoprene waterproofs:1.2) with yellow belt fleeing in the deep grassy bog gets trapped and (sinks in deep mud:1.2) struggling violently she is gradually pulled deeper and deeper into the deep mud , partial profile

It's not exactly literature and there's a lot of repetition but repetition strengthens the effect of prompt terms.

The associated negative prompt is:

extra fingers, extra limbs, Chinese, Asian , water, logo, deformed limbs, laying down, leash, hood, mask, cap, headband, headgear, hat, scarf

Next you need an image for ControlNet, I tried with one of your White Quicksand set pictures of O-Girl - wmbc22888e.jpg, there are several in this set that make excellent images for ControlNet, the plain backgrounds are perfect. ControlNet doesn't like busy backgrounds. Note: I used Canny as the Control Type.

Thirdly you need an image for ReActor, I'd use the same image as above, the face is right for the pose so should work well.

You may have to generate a few times to get a pleasing image and fiddle with weights, denoise etc.

I hope you don't mind me posting this, I know the source image is yours but assumed it would be OK as I am trying to help. When I did this using the images and prompts above I got:

Attachment removed.

The second method is to use Img2img with ReActor, you don't need ControlNet this way as the pose comes from the original image.

You need to use a good prompt/negative prompt again - use the ones I supplied if they are working for you.

You also need an image as base model, again you can use the same one - I did when I tried it.

You also need an image for ReActor, again the same image is a good one to try.

You will also have to fiddle with the denoising strength, too low and it simply copies the original, too high and you end up with something different altogether. I found, in this instance .6 kept the outfit reasonably recognisable (the belt symbol changed) but gave enough change to the background.

This all gave me the following result:

Attachment removed.

I hope this proves helpful and allows you to move forward in usage of AI. If you have any questions or need further explanation I'd be happy to help.

[MadMax359]

Jim will be the final judge... but i'm glad to see O Girl out in the wilds!

[cerberus]

Jim will be the final judge... but i'm glad to see O Girl out in the wilds!

yes indeed, Jim will be the judge regarding these images. The thing with AI generated images is that, at the moment you can't copyright them unless substantial work was done to customise the image. So, as far as the O-Girl images here are concerned, the only person who has any claim to the copyright is Jim, because these images are based on work he did or owns the rights to. Therefore, as far as I'm concerned, these images belong to Jim and he can do with them as he wishes, I make no claims on them.