YoinkoorAnimeXL

This model is an ensemble with various models(10+) on various weights and methods, this is the first version of it.
This model is great at anime style art. But cannot produce .

It works well with LoRAs and ControlNet while maintaining style at higher weights.

There are two models available for download, the FP32 model which is ~12.7 GB,

An FP16 version which is ~6.4 GB, loads faster, but comes with minute variations in details compared to the unquantized model. (might be less accurate)

Try it out here without downloading:

https://tensor.art/models/628619084524845515?source_id=nzy-oFjhnUmzp_cob3v88xki

Automatic1111

Use hiresfix x1.5 latent upscale 1.5 and high denoise 0.45-0.65.

comfyUI

Use the sdxl_refiner+hiresfix workflow to make the outputs 10 times better.

Though, I did not upscale in most of the examples.

Also try to use danbooru tags when proompting.

Share your results here, I will feature them if they are better than the current cover.