[Stable Diffusion] Prompt Sharing and Learning Thread

[devilkkw]

I suggest comfyui and dedicated model like wan2 1, ltx.
A search for those in Google for get many tutorials on how to setup and use it.

 

[JhonLui]

Hi, does anybody have problems with the ne Wai-Nsfw-Illustrius-SDXL v13.0?

V12.0 was working like a charm in Fooocus with Hyper, this one garbles up...
maybe I've messed something installing also Ruined Fooocus, Idk
Shall I reinstall it?
All in Stability Matrix by the way...
Thanks

Since we are here.. I also installed the latest Dvine checkpoint for Illustrious, and I have to say it totally delivers.

 

[Jimwalrus]

Hi, does anybody have problems with the ne Wai-Nsfw-Illustrius-SDXL v13.0?

V12.0 was working like a charm in Fooocus with Hyper, this one garbles up...
maybe I've messed something installing also Ruined Fooocus, Idk
Shall I reinstall it?
All in Stability Matrix by the way...
Thanks

Since we are here.. I also installed the latest Dvine checkpoint for Illustrious, and I have to say it totally delivers.

I have had one or two instances of it ignoring a LoRA (I'm using Forge). A restart didn't help, so it may well be the Checkpoint - I didn't consider it at the time.
EDITED TO ADD: I also noticed it started to create garbled output too, just clouds of coloured noise.

 

[JhonLui]

I have had one or two instances of it ignoring a LoRA (I'm using Forge). A restart didn't help, so it may well be the Checkpoint - I didn't consider it at the time.
EDITED TO ADD: I also noticed it started to create garbled output too, just clouds of coloured noise.

Thanks, than I'll go back to V12.0 and see what happens.
-edit- than it must be an error in the baked Vae..

 

[JhonLui]

.

Back to V12.0 back to perfect, thanks for sharing exp.

 

[aurelion44]

I am a super noob: what is the current state of the (free) art? i'm running comfyui with/on stablediffusion and using different checkpoints (dreamshaper, sdxl, and some nsfw versions). is this the best it gets for local, free image generation? I read the models/checkpoints for my current version somehow all use stablediffusion 1.5. the lowest of the all. is that true?

I also read we were waiting for some chinese local models without restrictions. in that context, why is wan2 not mentioned? is it because it apparently only can do text to video?

EDIT: Okay apparently over civitai I can chose the base model. Why are most recommended models here based on sd1.5 when sd3.5 large also exists? latter has way higher "score" on the elo board.

 

[Jimwalrus]

I am a super noob: what is the current state of the (free) art? i'm running comfyui with/on stablediffusion and using different checkpoints (dreamshaper, sdxl, and some nsfw versions). is this the best it gets for local, free image generation? I read the models/checkpoints for my current version somehow all use stablediffusion 1.5. the lowest of the all. is that true?

I also read we were waiting for some chinese local models without restrictions. in that context, why is wan2 not mentioned? is it because it apparently only can do text to video?

EDIT: Okay apparently over civitai I can chose the base model. Why are most recommended models here based on sd1.5 when sd3.5 large also exists? latter has way higher "score" on the elo board.

To answer your last question - I'd personally recommend Pony or Illustrious, precisely because we are mostly looking for good adherence to prompts of an NSFW nature.
Not had much contact with SD3.5, but AFAIK it struggles if you try to go NSFW. I'll probably not lose any money on a bet that asking SD3.5 for 'spitroast' would give images of pigs over a fire!

Flux is my preference for SFW, but that's mostly due to its excellent ability to duplicate celebs and the substantial number of LoRAs and alternative checkpoints available from the open community. Some of those Checkpoints do seem pretty good at NSFW too, albeit with their own quirks.

 

[modine2021]

is it possible to use a FLUX lora with WAN2.1? cause right there's just aint enough WAN people loras that are interresting except for one (Zendaya) i need more ethnic ones. Flux has a million

 

[dayloppoly]

Anybody have any tips on creating blowjobs and/or cumshots with ReActor? The ReActor plugin erases whatever is in front of their face.

 

[osanaiko]

Anybody have any tips on creating blowjobs and/or cumshots with ReActor? The ReActor plugin erases whatever is in front of their face.

Use Img2Img to add the sex parts.

1. make the base image with the celeb face using reActor

2. cut and paste in the dick and cum. Doesn't need to be particularly good job.

3. run the edited file through img2img. Use LORAs for to blowjob if that helps. use inpaint to isolate just the dick area if you find the rest of the image changes too much. mess around with prompt and settings, learn as you go.

(4. don't do this using the image of regular people/non-celebs. that's dickhead behavior)

 

[dayloppoly]

Thanks for the tips! I'll try it out.

 

[kaamist]

After Automatic 1111 Notebook for SD 1.5, with just some prompts and inpainting, which is the better alternative to go in SDXL ??

and I just love thick Big boobs of celebs , How to bring it like simply using inpainting..

I use colab and Sagemaker as of now by running Gutris notebook.. Any suggestion or recommendations ?

Thanks

 

[Sepheyer]

A nice step in video-generation with about 10 MB VRAM.

First, the vids:

You must be registered to see the links

Second, the tools:

You must be registered to see the links

Pretty insane.

 

[tarzan999]

A nice step in video-generation with about 10 MB VRAM.

First, the vids:

You must be registered to see the links

Second, the tools:

You must be registered to see the links

Pretty insane.

Those tools seem a bit complex. It's a lot more difficult than 'it's just prompts'.

 

[modine2021]

Those tools seem a bit complex. It's a lot more difficult than 'it's just prompts'.

what i use to make my 25 sec clips (done in 2 hours on 3060 12gig) it uses Hunyuan modal(creator working on adding Wan2.1) and also Hunyuan Loras

You must be registered to see the links

standalone

https://f95zone.to/threads/ai-art-show-us-your-ai-skill-no-teens.138575/post-17072113

 

[JhonLui]

what i use to make my 25 sec clips (done in 2 hours on 3060 12gig) it uses Hunyuan modal(creator working on adding Wan2.1) and also Hunyuan Loras

You must be registered to see the links

standalone

https://f95zone.to/threads/ai-art-show-us-your-ai-skill-no-teens.138575/post-17072113

Also SwarmUI should be a user friendly alternative for text to video (I didn't test yet).

Besides, did you try to overclock the memory frquency to see if it makes any difference in speed?

 

[Synalon]

Adding Sage Attention can help increase speed if it didn't install automatically you may need to install it manually.
When I use Framepack I have to install it manually, and I had to install it manually on Comfy when Hunyuan and Wan 2.1 first came out as an example.

 

[mosnew]

Just found this thread. I have been using SD forge and recently started inpainting to get multiple characters in the pic. Can someone help me with the best approach to it. i am facing issue with changing the character having sex, I tried Segment anything but masking exactly isnt working since the new character might not fit perfectly so i have been using a larger area making the mask myself but there are still issues where characters dont interact blend properly. I have watched multiple inpainting videos but obviously it doesn't help as the videos are shared from perspective normal images and singe character and dont consider when characters are lets say intricately invloved? any suggestions. I am use WAI-NSFW model and in inpainting using Euler A as sampler with 36 steps and 0.7 denoise strengh. Any inputs would be appereciated

 

[osanaiko]

Multiple characters with separate prompts for each might be better achieved using Regional Prompting, or a combination of regional prompting and inpaint masking.

I installed the following extension via the extensions function of Automatic1111, other webui's might have other ways to get it or something similar. I can confirm it works fine with SD1.5 and SDXL derived models and LORAs.

sd-webui-regional-prompter

You must be registered to see the links

I haven't tried to get it working with any of the Controlnet extensions if you're trying to use those for pose control etc.

 

[mosnew]

Multiple characters with separate prompts for each might be better achieved using Regional Prompting, or a combination of regional prompting and inpaint masking.

I installed the following extension via the extensions function of Automatic1111, other webui's might have other ways to get it or something similar. I can confirm it works fine with SD1.5 and SDXL derived models and LORAs.

sd-webui-regional-prompter

You must be registered to see the links

I haven't tried to get it working with any of the Controlnet extensions if you're trying to use those for pose control etc.

Thanks I have tried the regional prompter but it really doesnt work, I mean its one thing to have sun, moon, boy etc. but to partition it and have two people having sex turn out exactly like you want is not happening, LORAS get mixed, clothing is all over the place etc, after a lot of trial and error I decided to try my hand at in painting and that gave me so much better results with so less efforts just trying to work out certain chinks like blur, merging and proper way to mask when two characters are intimately involved. However If you do have any good guide for regional prompter for handling two people have sex better do guide me