|
Cenzontlecantor (273)
October 15, 2022, 8:03 pm
|
Posts: 60
Joined: Feb 2021
Reputation: 273
Liked by 13 members: Spongechu123 (Dec 23, 2023), CandyCornBandit (Feb 3, 2023), Bellylord123 (Jan 25, 2023), alter42 (Dec 22, 2022), Eskar13 (Nov 10, 2022), bigbellylover6929 (Oct 29, 2022), justthatguy (Oct 21, 2022), Chamilio (Oct 20, 2022), smashedpotatos (Oct 19, 2022), Fyra yu Chan 1999 (Oct 16, 2022), (Oct 16, 2022), colt3654 (Oct 15, 2022), pregnoid (Oct 15, 2022)
|
mrperson4321 (453)
October 15, 2022, 8:45 pm
|
Posts: 283
Joined: May 2020
Reputation: 453
October 15, 2022, 8:45 pm by mrperson4321 (453)
Anyone know how to make this art? Like what software to use? Or if this even is AI or if it’s too good to be true, because HOLY SHIT lmao
|
pregphil (3648)
Staff Member October 15, 2022, 11:05 pm
|
Staff Member
Posts: 760
Joined: Jan 2012
Reputation: 3648
October 15, 2022, 11:05 pm by pregphil (3648)
Are you sure this actually AI-generated @ Cenzontlecantor? If so, that's pretty remarkable. These look like drawings to me, but if anyone can confirm or comment, it would be appreciated.
|
User 70749 () October 16, 2022, 12:16 am
|
October 16, 2022, 12:16 am by User 70749 ()
It's AI-generated. AI struggles to produce anatomically correct hands, as would most non-artists. I would guess that eventually this will be addressed, either in the model or with a secondary AI hand-fixing tool.
Enthusiasts have trained an AI weighting model using the Danbooru archives as a dataset, an anime-style art collection which has highly detailed tagging and that helps AI immensely, it's called Waifu Diffusion. There's another derivative and fine-tuned weighting model that was leaked from NovelAI which is also specialised for anime-style art. I don't have these models as I don't run my own implementation of Stable Diffusion, but they're publicly available. AI models can also be merged and/or used simultaneously, and get the best of both.
If you want to create AI art you can either use a paid service (most of which give free trials but will filter nsfw content), run it on your own computer (best done using Nvidia RTX cards with a minimum of 4g of vram), or use a cloud service to run an implementation on a GPU-equipped virtual jupyter machine (which is actually easier than setting it up on your own PC, there are ready-to-go templates to help you do this). There's also an easy to use app for people with an Mac that has the M1 or M2 processor. There are easy to use versions and highly customizable versions.
It's not necessary to have a GPU to run Stable Diffusion, but it's around 10x slower when using a CPU instead of GPU.
The tech can also make photographic versions of art, and artistic versions of photos. If you can sketch a concept of what you want to see it can fill in all the details.
You can even train your own weighting model on pregnant content. The AI Teens section has examples of that by @ LadyBabyloin, for an example of what is possible today and a description of how it's done.
|
Cenzontlecantor (273)
October 16, 2022, 7:16 am
(Edited)
|
Posts: 60
Joined: Feb 2021
Reputation: 273
October 16, 2022, 7:16 am by Cenzontlecantor (273)
(Edited)
(October 15, 2022, 11:05 pm)pregphil Are you sure this actually AI-generated @Cenzontlecantor? If so, that's pretty remarkable. These look like drawings to me, but if anyone can confirm or comment, it would be appreciated.
(moved to the AI section pending confirmation) As I know, yes, is art of pregnant anime girls fron an AI.
Ahh, I didn't see we have a new section in the web destined for AI arts (sorry for the disorder)
WOW, this Is a very new thing to me.
It really take with the low guard.
|