Teds Pics

User avatar
BobDarkroom
Posts: 485
Location: Belgium
Contact:

Re: Teds Pics

Post by BobDarkroom »

LoyalFan wrote: Sun Sep 28, 2025 5:55 am
BobDarkroom wrote: Sun Sep 28, 2025 4:23 am

I have a question. I started experimenting with AI a few weeks ago using Comyui and did some trials with flux models recently. However, I couldn't generate NSFW (i.e. adult) results with flux models and according to some google search results these models cannot be used for that kind of images.
Browsing though your posts, I noticed you mentioned Flux models so I was wondering how you managed to make adult images with those models. As said, I"m a newbie so I may have missed something.
Thanks in advance for any hints!
Bob

The matter is quite complex, and you can find more info for example on YT searching with terms like "NSFW AI models".
In brief, there are NSFW models and LoRA's that allow you to create that kind of content, like for example:

https://civitai.com/models/667086/nsfw-master-flux


Hope it helps
Hi Loyalfan,

Thanks for your reply and the tip about LoRA.
I use models from civitai.com and with SD1.5 and SDXL models from civitai.com I have no problem with generating NSFW images but when I try using exactly the same text prompts on Flux models it doesn't work.
On the net I found following statement:


Flux models are not inherently designed for creating NSFW (Not Safe For Work) images,

However, I just found out that are workarounds by applying LoRA files like you already suggested so I will look into that. It seems there's also a security setting that needs to be changed to disable restrictions on flux models.
Anyway, I only started looking into AI image generation about 2 weeks ago so I'm still at the bottom of the learning curve ;)

Thanks for your hint
👍
Ted
Posts: 63
Contact:

Re: Teds Pics

Post by Ted »

I pretty much use flux exclusively now (running locally on my machine), and I haven't had any issues producing NSFW images. However the base model (flux1-dev-fp8) is "poisoned". That is nipples etc are corrupted. There are a few ways to get around that issue.

1) Use a different flux model. fuxCapacityNSFW, fluxedUpNSFW or fluxctasy are ones I use, altho any flux model that's trained as opposed to merged has a better chance of working. The quality of the images isn't quite as good as the base model, but at least the nudity mostly works .

2) Use LoRAs. I don't like using LoRAs, as I find they subtract as much as they add. Having said that the above images use a LoRA that I made to produce the whipmarks.

3) Generate images with the base model and then upscale them with one of the other models (I use fuxCapacityNSFW) The initial image will have corruption, but the upscaling will correct it. This is the method I now use, and the above images were created using this method.

I hope this helps.
User avatar
BobDarkroom
Posts: 485
Location: Belgium
Contact:

Re: Teds Pics

Post by BobDarkroom »

Ted wrote: Sun Sep 28, 2025 5:25 pm I pretty much use flux exclusively now (running locally on my machine), and I haven't had any issues producing NSFW images. However the base model (flux1-dev-fp8) is "poisoned". That is nipples etc are corrupted. There are a few ways to get around that issue.

1) Use a different flux model. fuxCapacityNSFW, fluxedUpNSFW or fluxctasy are ones I use, altho any flux model that's trained as opposed to merged has a better chance of working. The quality of the images isn't quite as good as the base model, but at least the nudity mostly works .

2) Use LoRAs. I don't like using LoRAs, as I find they subtract as much as they add. Having said that the above images use a LoRA that I made to produce the whipmarks.

3) Generate images with the base model and then upscale them with one of the other models (I use fuxCapacityNSFW) The initial image will have corruption, but the upscaling will correct it. This is the method I now use, and the above images were created using this method.

I hope this helps.

Thanks for your input Ted, it sure helps.

After reading LoyalFan's message I realized I didn't download the flux models from civitai but straight off the black-forest-labs site so that might explain why I couldn't generate nsfw images with those models.
I'll give it another try with flux models from civitai.
LoyalFan
Posts: 14
Contact:

Re: Teds Pics

Post by LoyalFan »

BobDarkroom wrote: Sun Sep 28, 2025 12:48 pm
LoyalFan wrote: Sun Sep 28, 2025 5:55 am
BobDarkroom wrote: Sun Sep 28, 2025 4:23 am

I have a question. I started experimenting with AI a few weeks ago using Comyui and did some trials with flux models recently. However, I couldn't generate NSFW (i.e. adult) results with flux models and according to some google search results these models cannot be used for that kind of images.
Browsing though your posts, I noticed you mentioned Flux models so I was wondering how you managed to make adult images with those models. As said, I"m a newbie so I may have missed something.
Thanks in advance for any hints!
Bob

The matter is quite complex, and you can find more info for example on YT searching with terms like "NSFW AI models".
In brief, there are NSFW models and LoRA's that allow you to create that kind of content, like for example:

https://civitai.com/models/667086/nsfw-master-flux


Hope it helps
Hi Loyalfan,

Thanks for your reply and the tip about LoRA.
I use models from civitai.com and with SD1.5 and SDXL models from civitai.com I have no problem with generating NSFW images but when I try using exactly the same text prompts on Flux models it doesn't work.
On the net I found following statement:


Flux models are not inherently designed for creating NSFW (Not Safe For Work) images,

However, I just found out that are workarounds by applying LoRA files like you already suggested so I will look into that. It seems there's also a security setting that needs to be changed to disable restrictions on flux models.
Anyway, I only started looking into AI image generation about 2 weeks ago so I'm still at the bottom of the learning curve ;)

Thanks for your hint
👍
You're most welcome! Something very important to keep in mind is that prompting works differently in Flux. For example, negative prompts have no use and parentheses are ignored.

This guy explains everything very well

User avatar
BobDarkroom
Posts: 485
Location: Belgium
Contact:

Re: Teds Pics

Post by BobDarkroom »

LoyalFan wrote: Mon Sep 29, 2025 5:16 am
BobDarkroom wrote: Sun Sep 28, 2025 12:48 pm
LoyalFan wrote: Sun Sep 28, 2025 5:55 am


The matter is quite complex, and you can find more info for example on YT searching with terms like "NSFW AI models".
In brief, there are NSFW models and LoRA's that allow you to create that kind of content, like for example:

https://civitai.com/models/667086/nsfw-master-flux


Hope it helps
Hi Loyalfan,

Thanks for your reply and the tip about LoRA.
I use models from civitai.com and with SD1.5 and SDXL models from civitai.com I have no problem with generating NSFW images but when I try using exactly the same text prompts on Flux models it doesn't work.
On the net I found following statement:


Flux models are not inherently designed for creating NSFW (Not Safe For Work) images,

However, I just found out that are workarounds by applying LoRA files like you already suggested so I will look into that. It seems there's also a security setting that needs to be changed to disable restrictions on flux models.
Anyway, I only started looking into AI image generation about 2 weeks ago so I'm still at the bottom of the learning curve ;)

Thanks for your hint
👍
You're most welcome! Something very important to keep in mind is that prompting works differently in Flux. For example, negative prompts have no use and parentheses are ignored.

This guy explains everything very well

Thanks for the link to the Flux tutorial.
Max-Plank
Posts: 5
Contact:

Re: Teds Pics

Post by Max-Plank »

Ted wrote: Sun Sep 28, 2025 5:25 pm 3) Generate images with the base model and then upscale them with one of the other models (I use fuxCapacityNSFW) The initial image will have corruption, but the upscaling will correct it. This is the method I now use, and the above images were created using this method.
Could you please elaborate on this? Is this an image-to-image method?
Ted
Posts: 63
Contact:

Re: Teds Pics

Post by Ted »

The upscaling is image-to-image. I can provide the workflows if needed.
Max-Plank
Posts: 5
Contact:

Re: Teds Pics

Post by Max-Plank »

I would be very grateful for the workflows. Is this done on a ComfyUI platform?
User avatar
BobDarkroom
Posts: 485
Location: Belgium
Contact:

Re: Teds Pics

Post by BobDarkroom »

Max-Plank wrote: Wed Oct 01, 2025 5:27 am I would be very grateful for the workflows. Is this done on a ComfyUI platform?
On the Comfyui platform there's a "Upscale image" node that you can connect between the VAE Decode node and the Save Image node.


Upscale node.jpg
Ted
Posts: 63
Contact:

Re: Teds Pics

Post by Ted »

Here are a couple of .png files. The workflow is included in them. You just need to drag and drop them into ComfyUI.

This one generates a lowres image :-
https://mega.nz/file/Ic13hRwR#sjbnl693L ... jeqxf1qNNA

And this one upscales it :-
https://mega.nz/file/pZVGTQiL#7htnr0-2d ... RPZFJNDRkE

And finally this is the LoRA that creates the whipmarks :-
https://mega.nz/file/gF8VwC6B#zshHazEpe ... JEpGMi8rSU

The other models used can be found on Civitai.
Post Reply

Return to “Art Contributions”

Who is online

Users browsing this forum: No registered users and 41 guests