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
New training process
FAQ
Comments (14)
Hello, is there a guideline for prompt words? I'd like to try it out and see how it works. I really like your examples.
No particular guidelines, use your usual prompt style. Training was done using joycaption prompts, so rather verbose and ~200-300 char length but other styles seem to work equally well. If you lack inspiration (as I often do) try copy pasting other prompts from other images and work your way from there.
Do post and link your results, I'm curious how people use the Lora.
It works perfectly with Qwen Edit 2511 Mixed and Lightning 4 Steps LoRa. It also works with Qwen Image 2512 in the GGUF 6K and 4K_M versions. I look forward to your new versions. Thank you for your work. ps: looks like it's not good making pen1s alone, only insert, so the head it's pretty bad.
So many of the Qwen loras around, but many don't specify if they work with Edit, Image, or both. Thanks for the info!
Thanks for trying it in so many enviroments. The Lora is trained on Qwen Image original, any other woking enviroments is pure unplanned luck. I'm training a Qwen Image 2512 version right now, expect it in a few weeks.
yep can confirm, this is great with qwen edit 2511
In all fairness, I can't differentiate between two holes either.
I can't wait for the 2512 version, although I suspect these are beta versions of 2512, o.2 is really good.
Thanks. The 2512 version is still cooking 😊.
@TimmyHodor I rent GPUs so if you need a hand creating LORA I may or may not be able to help. An a40 makes a 4hr 3090 local job into a 15 minute til download ready. I use AI toolkit, which includes qwen presets for 2512.
Thanks for the help offer, but I like my local setup. I cannot give you an ETA yet, but in the meantime, go checkout the competition. Qwen4Play released quite a good iteration recently.
works with qwen image edit 2511 nvfp4... "Convert the image into a realistic style" with no animeToReal loras. Also use 8step lightning lora CFG1.5.. NSFW is maintained and image turns pretty realistic
One of the best NSFW aid for Qwen Edit 2511 I've used so far. Excellent prompt cohesion and anatomy understanding out of the box, the best generalist for working out compositions by far. v0.2 is a major improvement, can't wait to see what is next!