MadMax359 wrote:half a boob is better than None!
But i would prefer the full view

Well, stay tuned there are more to come, but it may be a few days yet as I seem to have a bit of a queue of thigs to post.
bogbud wrote:
I love that teasing! Also your girls wearing leather or protective gear add a lot of realism to the pics.
Glad you liked that one, I thought it was quite cute. Well, regarding the protective gear, in that set I specified "multi-coloured bikini", SD seems to lake that rather loosely. In the past thought I've done sets with waterproofs and waders etc, it seems SD remembers those and adds them in for variety.
Duncan Edwards wrote:You can probably tell I'm not much of a wardrobe guy. How are you getting such interesting attire? Are you specific with it or do you just ask for "leather bikini" or something like that? Fascinating stuff.
I tend to do a mixture, sometimes I am specific, other times I don't bother and just see what I get. SD seems to take what you say with a varying degree of latitude anyway. If you don't specify clothes you get lot of nude, but putting in any clothing will mostly cause SD to pad the rest out. The more specific you are the more chance you have of getting what you want. It seems to understand most of the terms I use. You can give it choices e.g. "green/red", this will result in some greens, some reds and a mixture. It may throw in some other colours as well. it also seems to use previous batches as a source of inspiration, I've done several batches with "waterproofs" etc as prompt words so they get liberally scattered into most non-specific sets. You can force SD to adhere to the prompt a bit more rigidly by upping the priority of certain prompt words e.g. "(quicksand:1.3)", this boosts the likelihood of them sinking. Additionally, there is an option labelled CFG Scale which can be raised to increase the priority of the prompt over "free interpretation". It also seems to know what to do with costumes (e.g. catwoman costume), I'm not sure how it develops, I suspect it monitors what you save and don't save and learns from that, but have no proof of this.
Typically, my prompts start off something like the ones below:
"slim beautiful pert blue-eyed woman aged 25, light blonde short curly hair, medium breasts, wearing black cocktail dress ..."
"25 year old beautiful blonde woman wearing high boots and a multi-colour bikini ..."
"blonde European woman explorer 25 wearing tall boots ..."
It seems that SD knows what a cocktail dress and bikini and has a pretty good idea what explorers wear.
If the AI tool you use has a Negative (or Anti) prompt box, make use of that as well, my standard anti-prompts is:
"(worst quality, low quality:1.3),water, smile, stick, track, path, bad anatomy, missing fingers, extra fingers, sitting, kneeling, logo, text, watermark, children, boots on hands, paddle, carrying"
I'm not sure it takes notice of all these terms, but I thought most of them were worth a shot.