better make a new game. it could have been the best fantasy porn ip.
I just realized that all images, that i upload here, doesn't contain any generation data, which can be seen via "PNG info" tab or via CivitAI scanner. F95zone converts all uploaded images in .avif format and erases all generation data through this process. How foolish i was
...
...
Last edited:
Based on frame from Arcane.
Also based on frame from Arcane.
You don't have permission to view the spoiler content.
Log in or register now.
Checkpoint: Hassaku 1.3 A; Model: KOD+NN; Txt2img+Upscale+ADetailer.
Mamaragan
Member
- Jun 9, 2019
- 480
- 1,807
What's better between WAI and Hassaku?
I use only Hassaku 1.3A or 2.1fix and Prefect Illustrious (very rarely). All i know is WAI and NAI (NoobAI) also good for training.
Mamaragan
Member
- Jun 9, 2019
- 480
- 1,807
goatwhisperer3
Active Member
- Jul 22, 2017
- 672
- 2,512
Stumpled upon this thread because I got the itch to create some ai images (and have no idea where to start or what programs to use. So i just searched for "ai program" in the F95 searchbar)
After reading the "I want to generate images on my PC, what i need for that and what should i do?" on first page I wonder: Screenshots show some mentions of "stable diffusion" in the program names.
Quick google tells me thats one of the ai_generator thingys one can use. But the "original" sites in question want me to create an account and subscribe to get different levels of tools. Basically the more u spend the more u are allowed to generate.
so probably a naive question on a pirate site but: If I follow the instructions here what do I end up with? Some offline version of a generator? A pirated version thats basically the real deal but without subscribtions? A lightweight version that can be only used for Sabia and nothign else?
(When it comes to ai and chatgpt and all that stuff I feel like my grandma holding a smartphone.....)
After reading the "I want to generate images on my PC, what i need for that and what should i do?" on first page I wonder: Screenshots show some mentions of "stable diffusion" in the program names.
Quick google tells me thats one of the ai_generator thingys one can use. But the "original" sites in question want me to create an account and subscribe to get different levels of tools. Basically the more u spend the more u are allowed to generate.
so probably a naive question on a pirate site but: If I follow the instructions here what do I end up with? Some offline version of a generator? A pirated version thats basically the real deal but without subscribtions? A lightweight version that can be only used for Sabia and nothign else?
(When it comes to ai and chatgpt and all that stuff I feel like my grandma holding a smartphone.....)
Quick introduction into stable diffusion AI's.
You can install some UI to use image/video generation AI's at home PC.
I use this two UI's:
You must be registered to see the links
(relatively easy to use, looks like it forbidden)
You must be registered to see the links
(not easy to use, but powerfull and got frequent updates)Then you need to download base model (checkpoint), there is a lot of this kind.
Now i mostly use this:
You must be registered to see the links
Checkpoint differs from each other
1) by "architecture": SD1.5, SDXL1.0, Pony, Illustrious, Flux, HiDream, Wan, Hunyan and more;
2) by their knowledge: some good with anime (like Hassaku), some good with realism (like CyberPony or PonyRealism);
3) also some good for training another small auxillary models (LoRA, DoRA, LyCORIS and maybe other, but forgotten);
e.t.c.
After you have installed UI and downloaded some checkpoint you can start generate images, but in addition you can download auxillary models (probably LoRA, DoRA, LyCORIS).
Auxillary models can provide checkpoint with knowledge about new concepts:
1) characters; 2) styles; 3) poses; 4) memes and many more;
Auxillary models can be also founded on CivitAi or another ai hub.
All this is free, but you need decent GPU.
Depending on how much your GPU has VRAM you can run:
4: SD1.5;
8: SDXL1.0, Pony, Illustrious;
16 and more: Flux 1.D, HiDream (all), (maybe WAN, Hunyan, i didn't test them).
Nvidia's GPUs better suited for this tasks.
Flux and HiDream are censored models (HiDream less censored).
SD1.5, SDXL1.0, Pony, Illustrious can do lewd images (in this thread images was generated via Pony or Illustrious checkpoints, models).
If you don't have decent GPU, then you can use online generators for free (daily credits) or for money (subscription or just buying more credits).
Home UI's more powerfull because they have a lot of extensions like ControlNet (allows to "mimic" reference images) or ADetailer (fixing faces) and more.
Also you can train your own auxillary models at home (free) or at CivitAi (relatively cheap, i trained all my models here) or checkpoints (costly, you need to rent or buy one or more GPUs from server's segment like Nvidia H100 or Nvidia A100).
---------------------------------------------------------------
I don't use ChatGPT, so can't tell about it)
goatwhisperer3
Active Member
- Jul 22, 2017
- 672
- 2,512
Thanks for the quick info. Lots of terms I never heard of and probably will need to look into once I grab some free time.
If I get this straight:
If my PC is decent (It plays space marine 2 on highest settings, whatever that reference might be worth for, and it dealt fine with my "lets experiment with blender" phase. So I guess it hits the mark)
Then with your instruction/tools I got all I need to experiment and create somewhat decent ai pictures.
And all the "register here to create images" websites are not a given necessaty at all but provide a convenience service like
-save GPU: so people with weaker machines can still create images by using their servers/machines/whatever instead of the own
-help with prompts and directions: Meaning their ai is already "smarter" than whatever I will start with.
So if I tell mine "Show me a person with fire red hair" (using actual text, prompts or whatever) theirs will have a beautiful lady with red hair, cause its trained and anticipates what I mean while my first attempts at home might result in some monstrosity with a literally flaming head.
Sorry if I somewhat derail the thread too much by asking all his stuff thats surely already explained somewhere else and I am too lazy/dumb to look for right now.
Mamaragan
Member
- Jun 9, 2019
- 480
- 1,807
Yes. It runs locally, so no need to pay a ridiculous fee to use a model you can download yourself. All the pictures on this topic is done purely with Stable Diffusion webui.
Yeah, nah. Each model have their own way of handling prompt. For exemple, Illustrious checkpoints use only danbooru tags for prompt.
Other model, like SDXL or flux, use a more natural approach.
Illustrious is currently the way to go if you want a simple model that handle anime pictures very well. It's perfect for beginner.
To take your example of a prompt. To show you a person with fire red hair, you'll have to prompt it this way :
masterpiece, best quality, 1girl, pale skin, freckles, red hair, long hair, fiery hair, close up, blue eyes, smile,Result :
Last edited:
Just testing prompts that not related with KOD, but using KOD style to generation.
big_shl0m0
Newbie
- Jul 13, 2023
- 88
- 173
A Lora of just Nomos artstyle would be very interesting I think it.