Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
Existing user? Sign In

Sign In



Sign Up

DuffMan

Soaked Member
  • Posts

    1,219
  • Joined

  • Last visited

  • Days Won

    1

DuffMan last won the day on July 5 2013

DuffMan had the most liked content!

About DuffMan

  • Rank
    Soaked

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

DuffMan's Achievements

  1. If the A.I. isn't already trained on something you have to be literal, instead of using a phrase, even if the phrase might be obvious to us. Something like "holding her crotch" might be more literal, but then the A.I. won't associate other pee related actions from the scene around the model, like people's reactions etc. The A.I. actually takes a lot of renders to get any urine properly involved, (or at least coming from her groin and not hands) so a prompt like "peeing her pants" or "wetting her pants" or "peeing self" might work a lot better than a literal prompt. Without additional stain or urine prompts most renders won't actually have any pee in them, but will have the models behavior, like crotch holding, posture, others noticing, etc. If you haven't already, check out the tutorial I posted at the top of this thread. Getting the A.I. to do anything you actually want, (and not ten things you don't want) is way harder than most people realize until they try it themselves. Even if you start with my example prompts and/or my tutorial and skip the hard part, it might take 25 renders just to get something relevant, let alone alterable/"edit"-able. Looking at some of the other comments here and in other places, people seem to think that this A.I. stuff is quick and easy, it's not. It's slow, hard to figure out, and not exactly cheap/inexpensive. You kind of have to be a scientist to get anything usable right now. Take a look at the other Pixiv/Deviant art Omorashi posts to see what literally 99% of people there have managed to render so far with Novel A.I., it's not even in the same ballpark. That's not a brag, it's just the truth. Listening to those who haven't used any A.I. yet reminds me of older folks when CG first started showing up in movies, "Oh they just do all that on computers..." Which drove me crazy as I have done work in CG and stop-go animation, and the CG was way harder and more time consuming. (not to mention expensive) But the second a computer is involved people think you just hit a button and that's that. This reminds me of when "Prince" called the internet a "fad" about 10ish years ago. Great singer, but also dumb as a rock. Calling tech a "fad", let alone vanilla first gen consumer tech a "fad", has literally been wrong every single time anyone's said it about anything since forever. It only gets better/faster/cheaper past the starting point, and as far as starting points go, I've seen worse. (Anyone remember "AOL" "netscape" and dial up internet?) Deputy, amma be 75% real with you... Since this wasn't the only whining you've done on these images, I took the liberty of looking at your uploads/"art" to see what a master of craft would make. I guess I didn't know what I was expecting, but all you have are some chicken scratch pencil sketches of furries that look like a toddler drew them, and reposts of "homar's" work (along with some other artist's work) reassembled as one long picture. To quote Deadpool, "Maximum effort!". Your "about" bio says "a steady stream of art", and you have literally less than 10 original images (in half as many years) that I would be embarrassed to put my screen name on. Those who live in glass houses...
  2. For the fantasy one's with dragons I used- {{{photorealistic}}}, {{realistic}}, {{real life}}, {{photograph}}, {{skinny blonde triss peeing in her white and gold fantasy costume}}, {{{{pee staining and wetting and darkening her costume groin}}}}, {{{attacked by a dragon}}}, {{{fearful}}}, {{scared}}, {{embarrassed}}, {blushing}, {crying}, in a cave Swapping Triss with other names like Yennefer, Ivy Valentine and a few of Sophitia.
  3. It might be easier and less time consuming to start with the textures for 2 models/combatants at a time. There's only going to be so many people in one fight. That way you can work on several versions of wetting severity for fewer models to start with rather than front loading half the work.
  4. You would need to use something more advanced that either allows training or you can train locally. Like using a leaked version of stable diffusion or something, since they censor renders on their servers. (and I think also track prompts and ban users) Keep in mind, this is first gen consumer tech. Almost everyone tends to judge the future of tech based on vanilla first gen versions, which is always a mistake. It's why tech like VR was seen with such active hatred at launch, because it wasn't already the matrix in it's 1st gen consumer version, or didn't have the legacy software already for as big a library on launch day, etc. Or deepfakes, which were called crap because the 1st gen was indeed crap, but literally 12 months later they were on 3rd gen that was better than multi-million dollar Hollywood studio CG face replacements. (see superman's cg mouth in "justice league")
  5. DuffMan

    New "Scenes" batch (Novel A.I.)

    New batch of "scene" attempts in Novel A.I.
  6. Previously I'd managed to get decent scenes/events working well enough on their own, and with some difficulty figured out how to get stains/wetting, but the hard part now is to try and mix the two. Too many prompts and the specifics start to get washed out, no pun intended. Getting them both to hit in one render takes a lot more tinkering and dice rolls than any of the two separately. Here's a new batch of what I've managed so far. and the rest...
  7. When you say "the prompts" I assume you're referring to the reference prompts from the pixiv tutorial link at the top of the page. If not you should check out the link. {{{{pee staining and wetting and darkening her costume groin}}}} This one will generate some localized "groin" wetness. Some of the images I posted only use this, others use "edits" (as shown in the tutorial) that lean into the prompts. It might take 10-20 dice rolls to get something good though, and I'm always fiddling with order/magnitude of the prompts as well.
  8. This should help. https://www.pixiv.net/novel/show.php?id=18609737
  9. Newest batch using my better balanced prompts. I can manage to get the prompts to render without all the misty haze pretty reliably now, so going forward I'm going to try and get the actual "scenes" to be better, instead of the characters just kinda posing with a scene vaguely happening around them. The first tests I did had decent "scenes" that better incorporated all the characters, but lacked all the good looking pee/wetting that would actually look convincing. Up until now it's been one or the other. The challenge will be to balance "content" and "scenes" without losing either.
  10. The important part of the update I just made is the "undesired content" changes, not sure how useful that is in stable diffusion though.
  11. DuffMan

    Novel A.I. Pixel Art Omo

    Trying some pixel art styles with Novel A.I.
  12. What A.I. are you using? Those newer one's look like Novel A.I. but the first one's looked like they might be something else. If you're using Novel A.I. and trying to get stains I've got a tutorial for that. (which I literally just updated) https://www.pixiv.net/novel/show.php?id=18609737 Don't know if it's much use for any other A.I. generators though...
×
×
  • Create New...