This is yet another realistic NSFW lora, hence the name. It applies a photograpic realistic look and kinda does it all : AI influencers, everyday fully clothed pictures, erotic pictures, sex scenes, different body types and skin tones.
v1.0:
Yarn v1.0 is here!
This version is based on Qwen 2512, and was trained during 110k steps on a huge dataset. After extensive tests, I'm quite confident this is amongst the best realistic NSFW Lora on Civitai.
Although it was almost exclusively trained on SFW and NSFW of woman-focused photographs, as well as B/G interaction, this Lora is very good at plain SFW realistic image generation.
This iteration is by far the most stable and can be composed with other Loras such as character Loras. All tests were performed using the nunchaku version of Qwen 2512, with the 8 steps Lightning Lora.
v0.3:
This is an intermediate training step of YARN based on Qwen 2512. It performs amazingly well, and is still improving with further training. Pictures often have a distinct "studio look", which I like. I'm even finding myself using the Lora for purely SFW generations.
Pros:
Very good at photorealism
Male and female anatomy are very good
Diverse faces and body types
Composition coherence is improved
Retains high quality when stacked with a character Lora
Cons:
Still issues differentiation between the two holes. This is a captionning issue (I used JoyCaption), which will not improve with further training
Nude men are often missing the shaft if not specifically prompted
v0.2:
I changed the training settings and ran ~x5 training steps. This results in great improvements: v0.2 is much more stable, gives good results at strengths 0.5-1.5 and above. The artifacts present in v0.1 have disappeared. Overall, in my very biased opinion, this is a serious contender for best NSFW Qwen Lora on Civitai.
v0.1:
The lora is a tad too strong at the moment, works best at lora strengths 0.3-0.8.
Known issues : It knows the downstairs anatomy quite well, but can't really differentiate between the two holes in prompts.
Description
Yarn v1.0 is here!
This is the first full-fledged YARN iteration, trained on Qwen 2512.
FAQ
Comments (14)
wow, the new one is quite good, though 4 step lightning has some trouble (8 step is great)
v1 is amazing. nice work.
I have been lurking for 3 years here and this deserves a comment.
Best endorsement ever :-) ^^
Nah dude the realism is uncanny! I refuse to belive its qwen. I mean how! Its too good dude! Also im using qwen with the lighting lora and it always have the halftone artifact but some how your lora fixes that isse so regardless its nsfw or sfw im always slapping your lora on.
Same here :-)
Please post a few images, i'd love to see how people use the model. This might orient how I train the next iteration.
@TimmyHodor Sure thing man. Will definitely be posting images, its just im on a 2060 burnt toaster so it just takes a little while to cook LOL
Wow, this is absolutely wild, this LoRA is just mind-blowing in terms of photorealism. Seriously, bro, amazing work!
Thank you so much :-)
where can I find the res_2s_simple sampler?
is it just res_2s sampler with simple scheduler?
I copied all your settings and mine do not come out like yours at all. im using qwen_image_2512_fp8_e4m3fn Unet
qwen_2.5_vl_7b_fp8_scaled clip,
qwen_image_vae
i have both this lora and the samsung lora set to 1
Yes, res_2s_simple is res_2s with simple sampler. Try out beta as well.
I use the nunchaku version, but I don't think it will make much difference in terms of quality. Did you use the lighning Lora? You should. I'm also using an improved upscaler VAE that x2 resolution.
So I’m using Draw things app on cloud computing and something about this Lora changed from v.2 to v.3 where it now gives me a blue screen then crashes on the last step of generation. I’m using a lightning Lora and 2512. The same thing happens on edit 2511. I just tested v1 and it’s giving me the same result. What’s odd is that it’s only this Lora that does it.
Interesting.... The only thing I can think of is the rank of the Lora, which is higher in the cases where you have black images. Can you try with another high rank Lora to confirm?
Amazing work, people are sleeping on QWEN 2512