Hentai Mix XL - Road to freedom
The intend of this model is to be free of the censorship of SDXL models and be free to create nsfw hentai images of all kinds! Right now model is limited to the actions available in the NSFW POV All in One Lora, however I will be using this model as base to train new loras and keep adding to the repertoire. Loras will also be available in case someone prefer them.
That being said, stay tunned for updates and suggestions, posts and comments are always welcome!
Version 5.0
This update shifted the style a bit, maintaining the hentai capabilities while adding cunnilingus to the repertoire. This model tends to generate nsfw images more often than it's previous versions. The compatibility with character LoRas was also improved.
For instructions of how to use properly, reffer to this article.
Suported sexual acts without the need of a LoRa:
Blowjob
Cunnilingus
Licking Penis
Handjob
Cumshot
Doggy Style (pov)
Reversecowgirl (pov)
Riding Cowgirl (pov)
Titty Fucking (pov)
Anal (pov)
Fellatio (from side)
Suported franchises without LoRa
Overwatch
Disney
(not) Nintendo (peach, bowsette, samus) <-- scary statement
League of legends (poorly without prompt aid)
Street fighter (poorly without prompt aid)
There might be more including anime ones, just haven't tested them myself.
Version 4.0
This update integrated the VAE so in-site generation is as colorful as it's supposed to, hardcore capabilities were restored while keeping the style and quality.
Version 3.0
This update made the model loose a bit of their hardcore abilities gaining a lot of style and support for 2.5D images, it gained the ability to create characters from famous franchises to an extend.
Version 2.1
This update focused on giving the model a personality and a consistent anime style, without loosing the hentai capabilities. This model is able to create colorful images with the right prompts.
Version 1.0
This is the original version, it has a lot of issues, if you have his version, please consider updating. If you don't have this version, I advise you to skip it as it is very rustic.
Description
FAQ
Comments (18)
Version 4 was released, I wrote an article to describe the process to of how it was generated and the changes it has.
https://civitai.com/articles/3426
Also, this version comes with baked in vae so in-site image generation is as colorful as it's supposed to! thanks again for all support and please share your work, the buzz I get from it I use it to keep improving these models!
TYVM!
Do you plan to try to add cunnilingus into this? To my knowledge though there isn't any cunnilingus loras out there for sdxl
I had the intention of training a lora about to later integrate it to the base model, however the quality of the cunnilongus material that I have found that is strictly anime is scarce. Most of it is either censored or bad quality! I haven't given up tho, I might have to use AI generated material form the training
Hello there! I just trained an sdxl model for cunnilingus, I will most likely add it to version 5 at due time. In the meantime you an use the LoRa. Have fun!
Is there a way to get it to do penetration? penis vagina
Yes there is! in the first batch I posted some examples! I will post some more examples in a few minutes
Alright, posted them!
Basically there are 3 positions I know about that it can make, they are all POV and they are
reverse cowgirl, riding cowgirl and doggy style. If you check prompts you notice they work better if you hint it using 1girl, 1boy, penis, then the position!
Besides penetration, it supports fellatio
Hi, wanted to say it doesn't work to me, i clic on it but it instant put the previous checkpoint i choosed
Image generation is currently bugged, hopefully it will be back soon, it's an issue with the server, not the model. Maybe try again later and if it doesn't work, let me know I will report it!
Nice work! But model seems to struggle more than others with multiple objects. Especially if you try luxurious room, to much attention goes there, even 2 girls loose attention. UPD. tried 1 girl, room with red curtains, full body and it struggles.
But amazing on solo images so far.
Yes! indeed, when the subject is too small it looses detail, specially in the face, so if the frame has to be shared, less room for faces causes them to degrade. I have no idea how to combat this, sometimes making the image bigger or using impain, but as of in general text to image I don't know what I need to do to improve it!
This happens with all the models I have tried on sdxl tho (I haven't tried many) Inoticed it when working with my lewd dungeon lora, which creates women in the background.
@kuritsutian197 Yes, but to lesser extent, see my last couple generated lewd images. Inpainting in sdxl is meh for now. Seems like overall limitation, was less noticeable in SD due to hires fix. Potential way to combat in case of your lora - make it single background object, like city for example.
this doesnt work for me :( none of the versions do in fact, with or without VAE. it produces very weird results, ms-paint like flat images with maxed out saturation on every pixel and absolutely no depth. VAE remedies this BARELY, you can at least tell an attempt was made, but it is still worse than something a child could doodle. no other model does this for me, i have no idea what the problem could be.
That is very strange, maybe you can post one of the images you generate so I can take a look at the parameters and see what's going on
@kuritsutian197 It should be posted on my profile, lmk if thats what u wanted
Same ,Other models work fine though
You can try now the new version, hopefully this one work for you >_<
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.