Sorry for the late reply, was away from home for a few days.

>>5187

Nah, she's going into the breeding harem - pile for now. ^^ ...however these two young maidens here have just been selected to help finish manufacturing the kings new sword. What an honour! :3

>>5188

>how might you modify this method for things shoved in the vagina? like red hot fire pokers, spikes, etc.

Basicly that's what I've been doing for allmost all of my pics in the pussystabbing thread. Just generate the girl in the pose you want, if you don't need her in pristine condition for a before-pic maybe add blood from the getgo via prompting. If neccessary use ayaswan Lora at very low strength (~0.3) for that (haven't tested the Illu version of aya yet, but I suspect it shouldn't be to different, given the same dataset was used to train it iirc). Also depending on the shape and size of the object you want to insert it may help to prompt for a "gaping pussy". Then in your external photoshopping program of choice you add the pokers or spikes or as in my case the bladed weapon in a new layer on top of the image, scale the object, place it and then slide the opacity in that layer down until the object you want to insert becomes translucent. This way you'll exactly see the girl and her girly bits underneath the object and you can more precisely place it and the erase the parts that are overlapping (i.e. inserted or partially hidden behind legs, asscheeks, whatever else). Then you reset that layers opacity to 100% and save it. Send that to img2img for the inpainting and just as described before with help of anytest you can make the inserted (added) object better match the style of the rest of the image. In case of bladed weapons I'd use the pussystab lora at weight 0.8-1 for this inpainting and promptwise just be descriptive. For round things just prompting for "object insertion" should work. For anything else I'd just use the ayaswan Lora again at low weight and if the object is close enough to something that has a Lora existing for it I'd try to use that to help. Also sometimes instead of using relatively high denoise I'd make multiple passes with smaller denoise levels (0.2-0.3), if the AI otherwise just won't cooperate.

In general you may want to inpaint the object in smaller chunks. For example the branding rod I split into 3 parts/separate inpaint runs: first was "male hand from out of frame holding wooden grip of iron tool", then once that part looked good came the middle section with just "branding rod <lora:Branding_PonyV1:0.7>" and lastly "Red hot glowing branding stamp on asscheek, branding rod, dark skinned, blood, guro, white smoke, branding stamp with signaturetext "EeL", <lora:Branding_PonyV1:0.7>". If you tried to make the inpainting for the whole thing in one go you'd get AI weirdness for sure. As you can see you'll have to try to come up with a good mix of descriptive prompts, a good setting for denoise and controlnet weight on a case by case basis.

>and do you think it would work well with things like needles piercing the flesh, or scalpels cutting into it?

yes, the same should apply here basicly I'd say. You'd first create the subject with the wound (crudely photoshop the wound(s) yourself and then inpaint it with ayaswan Lora at 0.3-0.5) with prompts like "deep fleshwound injury (on bodypart x), bleeding, cut skin, fatty tissue, raw meat, guro, gore".

Btw, from testing it in conjunction with the pussystabbing lora it seems the branding Lora can help to make other things glow, too. Seems moderately gacha for txt2img with about 30 - 50% glowing sword output. But it works even better (as in the pic here) if you again do some of the work beforehand and add some glowy colors before doing inpainting.

>>5219

Super precice facial expressions I find hard to achieve, too, tbh. Like in this pic, her eyes rolled down; had to manip and inpaint that. Just prompting for it doesn't work (at least not with the models I'm using). In general I've found some (pony-) models really have a hard time with any more complex facial expressions anyways. So if you can't get what you want at first, switching models and inpainting might be worth a try.

For something more intense maybe try: (crying, tears, clenched teeth, tense facial expression, unimaginable pain, nervous sweating, panic, constricted pupils) or (crying, tears, mouth open wide, screaming in pain, tense facial expression, unimaginable pain, nervous sweating, panic, horrified, constricted pupils). Also sometimes "she is beeing killed" or "desperate" give me interesting results.

Most likely you've seen this, but just in case you haven't: https://civitai.com/articles/5492/danbooru-tags-complex-facial-expressions-for-ponyxl-autismmix