AI vs Sims

[poruan]

These are my three wives in the games.?

 

Spoiler

Spoiler

Spoiler

 

[Guest]

Spoiler

Version 1-2

 

Version 3-4

 

Version 5

Edited December 6, 2023 by CookieMonster901

[Dio_Wolf49]

I ran a few of the pics I took from my Top 10 Sexiest Goth Girl Outfits project through the A.I. and here is what I got.

 

 

 

Edited December 9, 2023 by Dio_Wolf49
Added 2 more Pics

[Taupeswift]

On 10/5/2023 at 3:44 AM, Nisiah said:

 

 

  Reveal hidden contents

 

 

  Reveal hidden contents

  Hide contents

  Reveal hidden contents

 

  Reveal hidden contents

 

  Reveal hidden contents

 

These are really impressive! What model pack do you use?

[Nisiah]

8 hours ago, Taupeswift said:

 

These are really impressive! What model pack do you use?

Stable Diffusion A111 with Realistic Vision model checkpoint

[llat229]

Anyone know a good place to figure the AI software out? It feels like one of those things where you mimic someones tutorial precisely, but they get this beautiful render while you are left with a mutant spawn.

[Venderin 2]

On 12/31/2023 at 6:00 AM, llat229 said:

Anyone know a good place to figure the AI software out? It feels like one of those things where you mimic someones tutorial precisely, but they get this beautiful render while you are left with a mutant spawn.

Hard to say. I've got more "normal" results, when I have written in commands like "human body", "realistic woman body", "human proportions", "human anatomy" etc. 
But other thing, is that it still from time to time gets messy, a specially when there are more than one person on the pic.

[Oops19]

On 12/31/2023 at 6:00 AM, llat229 said:

Anyone know a good place to figure the AI software out? It feels like one of those things where you mimic someones tutorial precisely, but they get this beautiful render while you are left with a mutant spawn.

https://civitai.com/ has a lot of sample images with prompts. You might want to look at them and copy the prompts to a1111. The result should match quite good if you have all needed checkpoints and lora models installed. Anyhow Detailer and Upscaler data is often not included and sometimes not all needed lora models are linked.

YouTube has a lot of videos to use A1111 or ComfyUI. With ComfyUI you may create more advanced workflows and split foreground and background to paint them again with a different noise. Use refactor to replace the face or to apply a face model which matches the sim. Use ADetailer to fix a broken face and Upscalers to improve the overall quality.

https://civitai.com/models/160207/sdxl-workflow-comfyui-realistic-skin-texture-portrait is a workflow to improve the skin, more or less.

[modella]

Spoiler

Spoiler

 

 

Finally got ahold of AI. I prefer the anime look~

[FranziskusFelix]

What an amazing topic! Here's my first-ish attempt. Gotta work on those prompts!

 

Spoiler

 

Edited February 15, 2024 by FranziskusFelix

[FranziskusFelix]

Another one of mine... btw, very good tutorial for prompts:

 

 

Spoiler

Edited February 15, 2024 by FranziskusFelix

[Kuntaman77]

My first try at converting sims to AI using StarryAI. [Just Portraits for now tho]

 

They came out pretty decently to how I envisioned the sims to look like

 

 

Spoiler

 

 

Spoiler

 

 

Spoiler

 

 

 

Spoiler

[Tried my best but I couldn't get the app to give her facial piercings]

 

 

 

Spoiler

 

 

 

Spoiler

 

 

[N0BLE117]

Is it possible to go backwards? I would love to add myself into my Sims game.

[Banks0902]

Edited May 22, 2024 by Banks0902

[Janssieve]

[Janssieve]

[jensonjests]

On 1/7/2024 at 9:46 AM, Oops19 said:

https://civitai.com/ has a lot of sample images with prompts. You might want to look at them and copy the prompts to a1111. The result should match quite good if you have all needed checkpoints and lora models installed. Anyhow Detailer and Upscaler data is often not included and sometimes not all needed lora models are linked.

YouTube has a lot of videos to use A1111 or ComfyUI. With ComfyUI you may create more advanced workflows and split foreground and background to paint them again with a different noise. Use refactor to replace the face or to apply a face model which matches the sim. Use ADetailer to fix a broken face and Upscalers to improve the overall quality.

https://civitai.com/models/160207/sdxl-workflow-comfyui-realistic-skin-texture-portrait is a workflow to improve the skin, more or less.

 

Necro thread I know, but I've been playing around with AI a lot and remembered this thread. Question,  when you say "Use ADetailer" are you meaning a specific tool, or do you mean a site, or something that downloads? 

 

I've been using mainly site based AI like Dezgo, Ideogram and Night Cafe, but have a rig ready for running it on desktop to do NSFW. For upscaling I used Leonardo if it is SFW, or a few other sites for NSFW (with mixed results) . 

 

Cheers!

[Oops19]

I mean https://github.com/Bing-su/adetailer which one can use. With ComfyUI one can create workflow and use other tools to detect image parts and/or to replace (inpaint) them.

 

[jensonjests]

On 8/1/2024 at 4:20 PM, Oops19 said:

I mean https://github.com/Bing-su/adetailer which one can use. With ComfyUI one can create workflow and use other tools to detect image parts and/or to replace (inpaint) them.

 

Ok cool, thank you. Have you used this one at all?

 

https://fooocus.co/

[Oops19]

ADetailer: yes
fooocus seems to be online only.
 

For ComfyUI there are so many nice YT videos, e.g. by Latent Vision. One can "play" hours with the settings

[jensonjests]

5 hours ago, Oops19 said:

ADetailer: yes
fooocus seems to be online only.
 

For ComfyUI there are so many nice YT videos, e.g. by Latent Vision. One can "play" hours with the settings

 

Fooocus has a download version too:

 

https://github.com/lllyasviel/Fooocus

 

And bit of info on it:

 

https://www.pcworld.com/article/2253285/fooocus-is-the-easiest-way-to-run-ai-art-on-your-pc.html

 

I'll play around with multiple desk top generators soon, once my PC gets upgraded. My question about Fooocus was if it does NSFW straight away, or you have to adjust it. And how it compares to the NMKD in regards to quality and accuracy. 

 

Is it possible to run both Fooocus and NMKD?

[Oops19]

Fooocus seems to be meant to be easy to use. Maybe more easy to setup than ComfyUI.
I'm quite sure that ComfyUI supports NSFW. There are many models on civitai.com to detail various body parts which can be used with ComfyUI (and/or other tools).
 

One can install many AI apps locally. While one might want to use only one at a time as the GPU memory and Cuda cores are limited. While one could configure one tool to use the GPU and the other one to use the CPU/normal memory to run two tools in parallel.

[jensonjests]

18 hours ago, Oops19 said:

Fooocus seems to be meant to be easy to use. Maybe more easy to setup than ComfyUI.
I'm quite sure that ComfyUI supports NSFW. There are many models on civitai.com to detail various body parts which can be used with ComfyUI (and/or other tools).
 

One can install many AI apps locally. While one might want to use only one at a time as the GPU memory and Cuda cores are limited. While one could configure one tool to use the GPU and the other one to use the CPU/normal memory to run two tools in parallel.

Yeah, I tried to dig a lil on Fooocus, it does do NSFW, but I believe you have to tweak it. I'll try it next week. I've only been playing around with AI art for less than a month, ComfyUI looks complex, but does Automatic1111 do NSFW too?

 

If i'm to understand it, does it come down to preference of UI as all of them can be tweaked by downloaded various models and LoRAs?

[Oops19]

a1111 and ComfyUI spport both NSFW.
ComfyUI comes with a simple workflow and as such is easy to use for simple images. There is no need to extend the workflow and/or to add lora's but one can do it later. As such I would avoid a1111.

[jensonjests]

23 hours ago, Oops19 said:

a1111 and ComfyUI spport both NSFW.
ComfyUI comes with a simple workflow and as such is easy to use for simple images. There is no need to extend the workflow and/or to add lora's but one can do it later. As such I would avoid a1111.

I know it is new, but Flux? This is open source and runs through ComfyUI or will be it's own thing?

 

 

https://blog.fal.ai/flux-the-largest-open-sourced-text2img-model-now-available-on-fal/