I don't speak English, so I'll keep it short. This is a NSFW SDXL model. Based on bigASP, NatViS: Natural Vision, LUSTIFY!, and my build that's nowhere to be found, thanks to the people who created these models.
Be careful with the hints, the model is very sensitive to NSFW. She constantly wants to do this... That's why she often generates funny situations))).
I'm not an expert in this, but I think something worked out, don't judge too harshly.
I have other models, if you like this model, I will post more.
Recommended settings:
Sampler: DPM++ 2M SDE/DPM++ 3M SDE,
Scheduler: Exponential/Karras,
Steps: 25-55,
Cfg: 2.5–6. I use 3.5 mostly
Hires steps: 10
Clip skip 2
Noise multiplier for img2img 1.1
ADetailer detector: mediapipe_face_full
Highres.fix: upscale by 1.5, noise reduction ~0.4.
For shots of people at a distance, you'll probably want to use highres.fix and/or adetailer.
You can try other settings...
Description
Bond 5 (NSFW). Model with VAE. I haven't tested this model much, I like the model, I will test it with you and post images. If the model is buggy, I will delete it.
About this version
Sampler: DPM++ 2M SDE/DPM++ 3M SDE,
Scheduler: Exponential/Karras,
Steps: 32-74,
Cfg: 2.5–6.
Hire steps: 10
Skip clip 2
Noise multiplier for img2img 1.1
ADetailer detector: mediapipe_face_full
Highres.fix: 1.5x upscale
You can try other settings...
FAQ
Comments (16)
Getting very weird results, overexposed with white "dots" all over the image. Same with V2.
You are quick! I love your first model, and I have your 2nd version which I haven't tested yet. And here you are with this 3rd model 5 days later, with the 4th model in the belly! :D Hahaha!
This is a fantastic model that is greater than the sum of its parts, imo. Really great quality, excellent lighting and backgrounds without having to prompt for any of that. Still testing, but prompt adherence seems nice as well. Very impressed so far, keep it up!
'Click a few times to let it warm up, but don't click too long, the model starts to do perversions' what this mean ?
but the important thing it can make potato with v3 ?
my pictures generate fine but for the finish it changes the color . some kind of light brown overlay dont know how to explain.
V4 seems like a great upgrade so far! Faces and especially eyes are so much better without the need for Adetailer in most cases
Impressive model, is there any database or list of the possible characters/celebrities it can recreate?
any suggestions about negatives?
Whats the difference with the V1.0_XL and the others?
I came to leave a comment about how I like the new XL, only to find XXX which is even better! At this point it is hard to describe the improvements, but XXX is a clear upgrade to me. Keep it up, great work!
"I'm not an expert in this, but I think something worked out" thank you for not being one of those people who even use AI to generate descriptions of their model lol, I love this honesty, and I believe you, I was so excited when I saw this update cos I've been loving v1, I could never get BigAsp to work for me but could see its potential and whatever you've done, it's working better.
thank you for your neat and short description , but i'm still have no idea what this good at , plus latest version means better ?!
генерирует смешные ситуации? это мы качаем!
damn , finally another model that uses natvis, by the way, would you say that the new alpha is an improved version of v2?
From my experience the vXXX_1.0 version is the best NSFW model, but other versions work worse.
However, I keep using the wrong clip skip value, 1 instead of 2, but I like results with 1 much better.