What is this?This is an experimental checkpoint I built by merging a very large number of checkpoints (sorry I didn't think to keep track of each one) using my own iterative merging workflow that involves repeatedly merging (1 step regular, 2 steps DARE) with randomized parameters and then picking the best candidate using a series of tests.The goal is a NoobAI checkpoint that is easy to use and works for almost any prompt, with a range of knowledge and styles, and providing a medium level of detail.<hr />Usage Tips:For the best balance between color and prompt accuracy I like a CFG between 4.0 to 7.0 and 22-32 steps.Use hi-res fix to produce larger images with much more accuracy and detail. I have DARE merged in weights from models designed for higher res images to assist with this :)For some example images I process multiple prompts independently and then concatenate the weights together (helps the model to avoid mixing unrelated subjects). I reconstructed those prompts in the image metadata using "BREAK" to denote where the prompts are concatenated, for compatibility with other workflows that use BREAK syntax.Use a mixture of danbooru &amp; e621 tags, and very simple sentences.

Natural Noob XL EPS (Anime/Furry/General)

Tuned to look a bit better than v2.0 with more accurate fleshed out details, and less oversaturated colors.For best results use a mix of booru/e621 tags and very simple natural language.vaemix v6 is now baked in, in place of Neptunia XL

naturalNoobXLEPSAnime_v30.safetensors

v3.0

v5.0

v4.0

v2.0

v1.0

Original Description:What is this?This is a rebuild of my other Natural Noob XL VPRED model, now based on NoobAI EPS and Illustrious models. A large handful of popular and obscure checkpoints and LoRas have been repeatedly merged into the mix using my own semi-random merging system and then I hand-picked the best result at each step using a series of tests to ensure a fairly consistent quality.The goal is a NoobAI checkpoint that is easy to use and works for almost any prompt, with a range of knowledge and styles.Usage Tips:For the best balance between color and prompt accuracy I like a CFG between 4.0 to 6.0 and 22-32 steps.For some example images I process multiple prompts independently and then concatenate the weights together (helps the model to avoid mixing unrelated subjects). I reconstructed those prompts in the image metadata using "BREAK" to denote where the prompts are concatenated, for compatibility with other workflows that use BREAK syntax.Use a mixture of danbooru &amp; e621 tags, and very simple sentences.

Files with SHA256: d346e35775598aa6c996d11057e2cc217eab12a100efdec15a04c28421270fcf