This LoRA model was trained on 100k images of normal to hyper sized anime characters. It focus mainly on breasts/ass/belly/thighs but is trained on over 47,000 unique tags.

In short, this model is a "fusion" of all of my other models. Hence fusion in the name. However if you really want the absolute largest sizes, it's best to use this model AND one of my specific hyper models. I've found that specialized models do work really well when they focus on one specific topic. I just wanted to have it all in one, so this is where we are.

Tag Info (You definitely want to read the tag docs)

Because hyperfusion is a conglomeration of multiple tagging schemes, I've included a tag guide "tag-data" in the training data download. It will describe the way the tags work (similar to Danbooru tags) and which tags the model knows best.
But for the most part you can use a majority of tags from Danbooru, r-43, e621, related to breasts/ass/belly/thighs/nipples.

The best method I have found for tag exploration is going to danbooru/r-34 and copying the tags from any image you like, and use them as a base. Because there are just too many tags trained into this model to test them all.

Tips

• If you are not getting the results you expect from a tag, find other similar tags and include those as well. I've found that this model tends to spread its knowledge of a tag around to other related tags. So including more will increase your chances of getting what you want.

• AbyssOrange does have a habit of putting nipples on bellies, try to avoid nipple tags when generating big bellies. use "navel" tags instead.

• Ass related tags have a strong preference for back shots, try a low strength ControlNet pose to correct this, or try one or more of these in the negatives "ass focus, from behind, looking back". Still if your prompt weights are too high you end up with front butt, lol

Extra

This model took me months of failures and plenty of lessons learned (hence v4)! If LoCon LoRA picks up in popularity, I may retrain it in the near future. I would eventually like to train a few more image classifiers to improve certain tags, but all future dreams for now.

As usual I have no intention of monetizing any of my models. Enjoy the thickness!

Training Hurdles

-Tagging-

The key to tagging a 100k dataset is to automate it all. Start with the wd-tagger (or similar booru tagger) to append some common tags on top of the tags scraped with the images from their source site. Then I trained a handful of image classifiers like breast size, ass size, hyper/not hyper..., and let those do the heavy lifting. Finally convert any similar tags into one single tag as described in the tag docs.

-Poor Results-

For a long time I was plagued with sub par results. I suspected maybe the data was just too low quality, but in the end it just ended up being poorly tagged images. Sites like r-34 tend to have too many tags describing an image like "big breasts, huge breasts, hyper breasts" all on the same image. This is not great for a model where you want to specify specific sizes. Using the classifiers I mentioned above I limited each image to a single size tag for each body part, and the results were night and day.

-Testing-

In order to determine if a new model is better than the last, it's important to have some standard prompts that you can compare with. x/y plot is great for this. Just keep in mind that the seeds between models will be totally different, and you likely need to compare dozens of images at a time and not 1 to 1. It's also important to compare new models against the base model output to make sure what you are training is actually having an overall positive effect compared to the origin model.

-Software/Hardware-

The training was all done on a 3090 in an Ubuntu docker instance. The software was Kohya's trainer using the LoRA network and lots of patience.