Another Damn Anime Model

SDXL fine-tune for Anime+ that's not illustrious/pony/noob based with 3D render & Traditionally media capabilities out of the box.

Buy me a coffee ❤

donations go towards funding more advanced open-source models.

My models will always be free but donations help cover the cost of more ambitious projects.

Changelog

2-15-25 ADAM XL v1.0 Release

Released ADAM XL v1.0

Usage Tips

Prompting

For V1.0, typical Danbooru tagging. Natural language and a hybrid prompting will be introduced in later versions.

Format of booru tags isn't strictly important, but I recommend:

artist tag(s), franchise/series/copyright tag, 1girl/boy/.., character name, everything else

Note: This model does not require quality tags, artist tags, or character tags.

Negative Prompt: This depends largely on the context of your positive prompt. Generally it's a good idea to keep the negative prompt simple, short and specific to your current needs.

Now, please ignore that I (mostly) used the same negative prompt in all my sample images 😉

"Do as I say, not as I do."

Sampling Parameters

There is no 'right' or 'wrong', aside from using CLIP Skip 1. But generally:

Sampler: Euler a | Scheduler: Automatic/Simple/SGM | Steps: 28 - 30 | CFG: 5 - 7
Sampler: DPM++ 2M SDE | Scheduler: Karras | Steps: 45 - 55 | CFG: 5 - 8

Training Info

(Coming Soon)

Limitations

(coming soon) But it should be stated that the models not perfect and will have limitations compared to larger heavily funded models like Pony, Noobai, and Illustrious