Another new day a new resumption

[MadMax359]

maybe here i'll call it the Max of Time....

and i'm betting it won't be the last, dear

that looks like hours of Incoming Tide Fun!

[cerberus]

That blonde woman is gorgeous! Good that she is in quicksand!

Not going to argue wit you there, she came out nicely.

Girls day out!
00007-2370462530.png

Wow, very nice and something you have not done before, so far. We need more of them! I guess they will have a lot of fun and can help each other (so they don't get stuck so bad like the one in the last pic of this batch)

I think I did some in the SD thread, but haven't for a while. This one was just an oddball thrown up by the AI, but I'll have to revisit the multiple subject prompts again, my stock of pairs are getting low.

Suddenly the cursed bog seethed and bubbled ferociously. She cursed it a bit more.
00150-576445787.png

Great area! Quite open and sunny but still desolate and lots of swamps-of-sadness-vides!

Yes, nice and atmospheric that one, I'll have to revisit that prompt for some more.

maybe here i'll call it the Max of Time....

and i'm betting it won't be the last, dear

that looks like hours of Incoming Tide Fun!

Probably not and she'll have to watch out, tides can come in really fast.

I'll have to be quick again tonight, site went down just as I was about to post, so it's a bit late.

00015-60953710.png

00022-1857391017.png

00074-1591682332.png

00078-419133648.png

00144-3635304631.png

[cerberus]

Rest.

00063-4072449354.png

00088-1800377040.png

00092-2863765151.png

00111-2019205121.png

00145-2179071134.png

[Scarol]

Nice photos! What changes the appearence of the Woman? Are that the samples steps or the seeds of GF scale, when I reload the Ui, and type the correct promps I get a different face.

[cerberus]

Nice photos! What changes the appearence of the Woman? Are that the samples steps or the seeds of GF scale, when I reload the Ui, and type the correct promps I get a different face.

Thanks. The main things that result in a change to the subject's appearance are the seed, the prompt and the checkpoint model. The seed identifies the starting noise pattern, different noise pattern different resulting image. The AI will use the prompt as a guide to the resulting subject but may ignore terms it it can't fit them with the noise pattern. The checkpoint models are trained using a set of images (usually thousands) so different checkpoint models will give different faces and generation results. Most of the other settings have a smaller effect but can shape the mood/feel of an image. In the case you describe, it is the change of seed that is most likely causing the variation in the image.

[Scarol]

Thanks, I saw that the most models has 10 Numbers, if I’m correct if I see a model and I choose the checkpoint (reallive for example) and I see the model (Karras for example) and the seeds (10 Numbers) and the GF scale, and I use the words sinking in quicksand in prompt then I will see that woman what I saw on the site will be the same what I will see? Is it obliged to put 10 Numbers on the seeds or is a 4 or 15 numbers also usable or are the result more bad?

[MadMax359]

oh, you meant "the rest"--- that wasn't a suggestion! little chance of resting with this batch

[cerberus]

Thanks, I saw that the most models has 10 Numbers, if I’m correct if I see a model and I choose the checkpoint (reallive for example) and I see the model (Karras for example) and the seeds (10 Numbers) and the GF scale, and I use the words sinking in quicksand in prompt then I will see that woman what I saw on the site will be the same what I will see? Is it obliged to put 10 Numbers on the seeds or is a 4 or 15 numbers also usable or are the result more bad?

I think you are misunderstanding the terminology I use; when I refer to a checkpoint model it's a mathematical model - a collection of algorithms and, characteristics and calculations that, when fed to Stable Diffusion along with the other settings, result in the generation of an image. The images on the website will have been generated using the related checkpoint model, but you won't necessarily get the same face if you copy the parameters for use with your prompt. That said, checkpoint models do result in a limited number of faces, so you will see similar faces appearing, especially if you specify particular hair colour etc. Seed numbers contain up to ten digits (apparently the maximum value allowed is 4294967295) but may be fewer, it is simply a number that specifies the starting noise pattern from which the image is generated. The CFG Scale (Classifier-free Guidance Scale) defines how closely the image generation process sticks to the text prompt, the higher the setting the closer it tries to adhere to the text prompt, though sometimes it struggles and you get artefacts and distortion.

[bogbud]

Rest.

00063-4072449354.png

Lol, no rest indeed!

Love her medieval braids and those tiny mudspecks (first i thought those were freckles)
Also a very nicely tailored rubber-trenchcoat. Too bad it does not help her staying dry or clean

[cerberus]

On another thread there's some discussion about facial expressions in AI images and the general assumption that photorealistic AI models produce boring deadpan lifeless images. I've expressed my thoughts there on this but some evidence would help. I don't wish to pollute the other contributor's thread with my images, so I'll post a few here and let people decide for themselves.

Does this really need amplification?

00155-2638294742.png

I think that look says "you will pay for this" pretty emphatically!

00354-2816950020.png

Could be a couple of emotions expressed here, I'll go with "scared shitless"!

00508-4213892721.png

To me this says "totally pissed off"!

00422-659038068.png

and this, well, God knows, I'll leave that to your imagination.

00395-1526347475.png