This merge is very aged and the complexity is high. This has some training for some of the earlier models, but it's primarily built to be structurally powerful and pose-strong. It's yielded the most consistent results of the merges I've mixed, so it's quite powerful.

It produces NSFW imagery more often than anything, but it's not pure NSFW. It can create images of many variations of many types.

It handles tags from E621, which function fairly well on Clip Skip 2 in comparison to the Yiffy originator. It handles danbooru, gelbooru, and r34 tags fairly well. There's a bit of inconsistency when working with some poses when using clip skip 1, but it's fairly solid overall.

Yiffy and furry requires additional power to the e621 tags that define style, so be aware it's a bit weak, but it can be used with a bit of stubborn elbow grease.

There are a series of negative prompt correctors, and positive prompt solidifiers that can produce solid output from this model. It's a bit rough around the edges without correction tools at times, but it definitely shines when there are some correction tools applied. The various poses, the diversity to the tag pools, and the solidity of the originating models make for an interesting experience no matter what you type in.

I lost track of all the models that I used to make ObsidianV2 and ObsidianV1 was combined to create Obsidian V2. There's a large chain here.

This is not meant to produce mature themed images for any specific real person.