About the model:

My goal with this mix was to create a mostly 2d anime style model that responds well to varied prompts, including a relatively wide range of existing characters without the need of additional LoRAs. NSFW was not a focus for this model, but it can still do it. I have done little testing on LoRA compatibility with this model because I primarily use style LoRAs and this model already has the style that I wanted it to have.

I'd love to see what you create with this model!


Prompting:

Kakigori responds best to booru style prompting, although it can handle some natural language prompting.


Example prompt:

kirisame marisa, blonde hair, side braid, waist apron, witch hat, hat bow, yellow eyes, superb, looking down, splashing in a stream, nature, outdoors

I find that a simple negative prompt without any negative embeddings has worked best for me.

Example negative prompt:

(Worst Quality, Low Quality:1.2), border, nsfw, skimpy, grayscale, multiple_girls, (watermark:1.2)

Recommended settings:

Clip skip: 2

Base Resolution: ~512x768 or 768x512

Hires. fix: 4x_fatal_Anime_500000K_G / Upscale by: 1.5+ / Denoising strength: 0.48

Adetailer: face_yolov8n

Sampler: DPM++ 2M Karras/DPM++ 2M SDE Karras

CFG: 7.5 / Steps : 20


Credits:

I would like to give thanks to the authors of the following models which were included in this merge. If you like my merge then I would encourage you to try these out as well.

AuroraONE by Luna

Toonyou - JP - Alpha 1 by Bradcatt

ExpMix_Line - V3 by Mods13

(VAE) kl-f8-anime2 by hakurei

v2 as above with the following additions

Awooooooo/Awoo7 by yoinked

vanilla by NoCrypt

Lilliput by Cinsdia

ssambateaLoconStyleV2 by AshtakaOOf

v3 as above with the following additions

amedira-bv3 by AshtakaOOf

Based67 by AnonymousM


Why baked VAE?:

Because I see too many people using models without the correct VAE too often, the impact to file size is tiny, and I don't expect this model to be great for training from anyways. You can still load whichever VAE you would like in WebUI and if you're more than welcome to prune the VAE yourself if you'd prefer that. The VAE I used is a customized version of kl-f8-anime2 with lower contrast than the official version. If you would like a copy of the VAE baked into this model you can get WD1-4-kl-f8-anime2-bless09 here.

This model may not be used on any paid generation services. You may not sell this model or any merges that contain it. This model is for personal/private use. Commercial use is not authorized.

You are more than welcome to include this model in merges as long as those merges adhere to the same license as this model. If you upload your merge all I ask is that you include a link to this model in the about section like I have done here.


license: creativeml-openrail-m

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference API
Unable to determine this model's library. Check the docs .