V3 Is HERE!
All images have WF embedded, I recommend you to use euler/beta 12 steps on turbo. It gives better skin and textures.
Description
V2 double the trouble.
FAQ
Comments (18)
Thanks for this, can you say what lora and strength layer bypass lora you have folded into this?
None, it don't need it
Just download any image and use the wf embedded
@alcaitiff Wow ok, so i guess any finetuning at all takes care of it by itself. Your v1 works really well for overall prompt following compared to the original turbo ,even not NSFW. I've been recommending it all over the place.
Nailed it!!! with V2 awesome stuff. Really top tier!
Good work ! I did some experiment to reduce lora/lokr size. if you want a "light" version of the lora ;) https://civitai.red/models/2742336/nsfw-krea2-experiment
Amazing work!
Keep it up with the gallery posts ROFL. Made my day.
This looks awesome but does it work with the base model or any other checkpoint too? What happens if you combine it with an NSFW checkpoint, does it become double triple NSFW?
i used it with a nsfw checkpoint and yes it can overcook the picture so i would recommend u to run it on the base model
@xvvrcv if the checkpoint mixed it on the merge it will be very bad.
"double triple NSFW", man, my very first SD 1.5 prompts were exactly like this. Just hundreds, to thousands of random words and phrases jammed at the end of every prompt. I didn't feel so bad when I learned that embeddings were just that without straight up hallucinating lmfao (I mean an embedding is giving weights to words via images to influence the CLIP, I was also influencing the CLIP... with nonsense)
Still, the answer is no.
The way to have the answer be yes is to lower the ranks on all the input loras such that their respective influences won't impact eachother -- they still will though. Generally you pick just one, or figure out how to balance their strengths (maybe one higher, a couple lower). You can maybe find a balance with many, but it's not easy without creativity lol.
What strength should i keep for this lora
1
I don't see any fields to choose the lora on the workflow that I got from Next diffusion, and while the model is unrestricted, after 3 attempts I wasn't able to generate a voluptuous big breasted woman, like the one in the simple gallery, although several references were included in my prompt (e.g. voluptuous, 36DD, big breasted)
The face wasn't good in the 3rd attempt with some deformities.
Any advice is welcomed.
What is Next diffusion? I'm sorry I'm lost here.
1. likely referring to some website that'll let people add keywords from loras.
2. If you're using the prompts provided in the gallery, you should be seeing similar results.
3. Use sentences.
4. Make sure to refer to things you likely feel you shouldn't have to: like lighting, or the setting, or the camera, or the action. Each little snippet lets the model build a whole idea for how the image should look -- it's how to get natural real images.
5. As per the actual question, the actual image at hand, since you didn't link the exact one it's hard for us to know. All the images have prompts. If you're finding you aren't getting the same result, it could simply be the model/generator you're using, or the OP got lucky with a seed. We don't know. I've written over 1000 characters and deleted them just for this one message. I want to help you. I don't know how bro.
I discovered that this Lora wasn't selected in the Lora node, in order to select it I had to deactivate Nodes 2.0 from the settings, but after selecting the Lora the whole things get's stuck. I have 24GB VRAM and 64 RAM, what am I doing wrong?