Sign In

Heroine Chic - A Hyperfused SFW/NSFW Semi-Real Base Model

114
973
23
Verified:
SafeTensor
Type
Checkpoint Merge
Stats
537
Reviews
Published
Nov 28, 2023
Base Model
SD 1.5
Hash
AutoV2
EF36899C92
default creator card background decoration
iamxenos's Avatar
iamxenos

V2 - Babes and Bimbos

This is a merge of the original Heroine Chic Hyperfusion blend and my XenoBABES model with some tuning and balancing to help bring it even closer to realism while improving on other factors, as well. In addition to taking to some of the more fun outfits and styles of the BABES 3.0 model, there are a few other fun things happening here.

  • The default female body, while more realistic than before, is also shaped more like a Western Comic style character - thin, yet curvy.

  • The trigger "Bimbo" makes a very specific bimbo look - with bolted on tits, tiny waist, and wide hips. Aside from hair potentially pushing toward blonde, maybe some heavier makeup, and a bit of angles in the cheeks, very little happens in the face area in order to preserve the original looks of your character. (You can often use things like "big lips" and "big eyes" to help give that classic Bimbo Barbie look. Calling your subject a "ThePit Bimbo" (thanks to Babes) may also further refine the bimbo style. Play around with it.

  • This model is now optimized for, and one of the recommended models for use with my original XenoBABES Characters.


Like it? Buy me a beer.

I'm using Civitai's trainer - so Buzz Point tips are helpful too - even just a few add up.

Want Tips, Tricks, Pix of Things We Can't Show on Civitai?
Join the XenoEngine Discord to discuss tips and tricks, complimentary Lora's and Embeds, and to post (and see) images not allowed here on Civitai.

You could also drop @Throwawayjim and @alexds9 a little love and buzz to keep up the great work there. I'd love to see that project continue down the road.


ORIGINAL NOTES AND BACKSTORY

The new version of Hyperfusion dropped this week - as a full checkpoint - and I was very excited. The earlier versions always gave incredible amounts of control over body shapes and sizes - but they always ended up pushing hard toward the anime/hentai styles it was trained on. Quite understandably, the new checkpoint works similarly. I also found the hands coming out hammy a bit more often than normal, and some of the art styles (at least with simple prompting) just felt a little off - lacking a little life or detail.

As great as it is, I wanted to try to make it take other styles - and go as real as possible without losing the body control of Hyperfusion. With mixing in some of art style knowledge of XenoENGINE Artstyle along with the semi-real aspects of XenoGASM, I ended up with this awesome little piece of magic.

If you download the CSV file that comes along with Hyperfusion - those 14K+ tokens should all be represented and generally work - though you will likely need to turn up the weight on many of them since it needed to be watered down a bit to take on the other styles.

This model also doesn't really have a fixed "default" style - so simple calls without styles can go all sorts of fun and interesting places. Like my XenoFLEX model - you'll need to call your styles and call your shots a bit more than with my other models if you want it to hold a certain look.

And also keep in mind - a lot of the unique tokens in there are ONLY ever trained on Anime/Hentai images - and when you use those and turn up the weights, they're naturally going to push back toward that Anime style. As such, some of the more obscure (but still totally fun and unique) tokens require playing a bit of a balancing game between the token and the styles you're trying to accomplish.

If you're going for a super real look and pushing real hard toward it - many of the more extreme sizes and shapes just don't seem to work - but someone may find a way to prompt it. It's in there - it's just sometimes hard to get out without it turning Anime. In another generation or two of this model, I'm hoping to even out that balance a bit more - without losing too much more on the function side of things.

NOTE: On several of the example images, I used my "Enhancing Reality" tips by refining and ADetailing with my XenoREALITY model (in the face only - the body tends to do funny things when Hypefusion Exclusive tokens are used, but I have some ideas for that down the road).