- May 26, 2018
- 33
- 297
Another anime-ish attempt that I liked even if some of the detail got a bit washed out.
- May 26, 2018
- 33
- 297
I liked the shadow and lighting effects on this one plus I wanted a full-body image. It's probably that I'm using bad prompts but I find it can be difficult to get the AI to move the camera far enough back, particularly if you prompt for any kind of specifc details.
- May 26, 2018
- 33
- 297
- May 26, 2018
- 33
- 297
I'll see if I can do some images of other Holiday Island characters in the future, although I'll have to train a new model or do a textual inversion to do that.
Anyway, as always all credit to the legend himself, Darkhound, for creating such awesome girls in such an awesome game and also, credit to those posters making fan art for this game by actually doing renders! I admire you guys a lot and appreciate your work.
Anyway, as always all credit to the legend himself, Darkhound, for creating such awesome girls in such an awesome game and also, credit to those posters making fan art for this game by actually doing renders! I admire you guys a lot and appreciate your work.
- Jul 3, 2019
- 7
- 9
- May 26, 2018
- 33
- 297
- Mar 17, 2020
- 263
- 259
- Mar 1, 2020
- 1,911
- 6,297
Heather's Valentine Photo Shoot.
Last edited:
- May 26, 2018
- 33
- 297
Yeah, I can do that if that's what people prefer. I usually generate 1024x1024 anyway, I just didn't want to spam the thread with lots of large images kinda like I did when I posted the Alice stuff.
I did make a couple of 1024x1024 images in that initial Jennifer batch which I didn't post earlier and I may as well share them now - Nothing too exciting, one is just a headshot and the other frankly I'm not totally happy with the accuracy of the face, plus both images are kinda similar to stuff I already posted.
- May 26, 2018
- 33
- 297
I meant to post this one earlier too...AI Alice in the rain, SFW I'm afraid.
- Sep 15, 2021
- 1,048
- 4,012
Safe for work maybe, but she'd still give me a heart attack!I meant to post this one earlier too...AI Alice in the rain, SFW I'm afraid.
Thanks! Although I wouldn't recommend using it, these were all done with NMKD as it's a very easy install (no messing about with Github, command line etc.).
The model was mostly F222, with some HassanBlend when F222 was struggling due to the 'topographical extremes' of Holiday Island!
They weren't actually deliberate face-swaps or inpainting, they were straight img2img with the influence set to 0.9 (not sure what that equates to in regular SD, probably the denoising strength) to minimise drastic changes. I didn't want to stray too far from Darkhound1's work.
I've started to use regular SD webui now and, thanks to some recent developments, Dreambooth no longer requires you to have a GPU with insane VRAM. My 3060's 12GB seems to be enough to do this type of training.
I'm testing it atm, one of the projects on my list is to train it on my earlier "face swaps" of the HI girls to improve consistency.
Where do I find these parameters?
- Sep 15, 2021
- 1,048
- 4,012
If you're wanting to have a go, bear in mind it's a real 'Chronophage' ("time-eater").
Basic starter info posts can be found in this thread, along with other useful advice for beginners and intermediate users. I don't think any of us would regard ourselves as advanced yet.
For the sort of NSFW img2img work involved in what I did, you will really need to be doing it on your own machine, cloud compute doesn't like that sort of thing. That does necessitate a fairly powerful PC, see the links above for guidance.
Basic starter info posts can be found in this thread, along with other useful advice for beginners and intermediate users. I don't think any of us would regard ourselves as advanced yet.
For the sort of NSFW img2img work involved in what I did, you will really need to be doing it on your own machine, cloud compute doesn't like that sort of thing. That does necessitate a fairly powerful PC, see the links above for guidance.
- May 26, 2018
- 33
- 297
Thanks for the links! I’ve been using SD for a few weeks now, but it seems like there’s still a ton of basic, entry-level stuff that I don’t know, so any resources are appreciated.
It actually never even occurred to me to search for a Stable Diffusion thread on this forum! <facepalm>
- May 26, 2018
- 33
- 297
YouTube is your friend if you want to get started with AI art generation, lots of excellent tutorials on how to get started.
At least that’s how I got into it. I honestly just blindly followed along with YT tutorials, not really knowing what I was doing.
I will agree with Jimwalrus 100%, it is highly addictive and can be a massive timesink. It’s just a ton of fun trying crazy stuff to see what you end up with.
For me, I also find a big part of the time expenditure comes through trying to fine-tune your art generation. I have learned that just generating any old AI art is very easy, but generating good AI art can be a bit of a challenge. I must have created hundreds of images over the last few weeks, but I still don’t think I’ve made one yet that was precisely what I wanted. You just have to accept that your creations are almost certainly going to have some artifacting or deformity or some other kind of AI weirdness going on and correcting can be more trouble than it’s worth. (Although this problem seems to be improving all the time as the tech develops.)
Also concur with JimWalrus, you will need a decent GPU and you don’t really want to generate images on the cloud. (VRAM is key – I would much rather have a 12 GB RTX 3060 than my 8 GB 3060 Ti)
My Alice and Jennifer generations were created used models I crafted with Dreambooth by renting time on a RTX 3090 on the cloud (runpod.io) but there are now, much more efficient ways to do things that don’t need 24 GB VRAM (I’m thinking of LoRA)
Anyway, there’s lots of online tutorials available if all this sounds completely foreign, and the links JimWalrus posted are a great start.
- May 26, 2018
- 33
- 297
Took Jennifer and Alice off of Holiday Island for a few days on a an AI field trip...
I've had a lot of fun getting to AI to put the HI in some "different" situations. SFW again I'm afraid.
I've had a lot of fun getting to AI to put the HI in some "different" situations. SFW again I'm afraid.
- May 26, 2018
- 33
- 297
