Introduction
Anima Cat Tower is Anima's fine-tuned model.
This model was trained with the goal of enhancing anime style.
Please refer to the official Anima Huggingface repo about Anima.
How to Get Started with the Model
You need to use the webui that supports Anima.
ComfyUI
Forge Neo
Positive prompt
masterpiece, best quality, highres, absurdresNegative Prompt
worst quality, low quality, blurry, jpeg artifacts, lowresLicense
Description
Anima-Preview based
v0.2 was trained with 1280px and 1536px bucket resolution. (v0.1 was trained with 1280px bucket resolution only). This reduces the noise that occurs during upscaling.
Positive prompts and Negative prompts have been revised (v0.1 is the same).
FAQ
Comments (11)
another's trained ckpt cant wait to try
its baked VAE or need VAE?
@saytolol12 VAE is not baked. You need Qwen-Image VAE.
It seems that catastrophic forgetting is already in place:
https://civitai.com/images/121086308
一番良い
I am quite fond of it
guys, anybody upscailing images? Please share the settings
do not go past 1.5 when not using a separate refiner model. I use various illustrious-based models as refiners to push past 2.0 upscale, with refiner ctrl% (init image creativity in some UIs) at 1, 1.5 rarely 2.0. If I want to stick as close to the original image as possible with minimal to no changes, I use the OG as an init image with creativity set to 0.
Bear in mind I use SwarmUI to do all this, so your mileage may vary when using other UIs, but I'm quite satisfied with my upscaled outputs. Bonus is that SwarmUI can also do adetailer work out of the box via an intuitive built-in segmentation feature.
@DevilSShadoW yes, I also used the Illustrious model for upscaling, and it turned out pretty good. But if I used anima, the upscaling was veeery bad.
Thanks a lot for your answer)
@DevilSShadoW I normally use ReForge (will of course have to use Forge Neo for this). Any tips for using the hires.fix for that? I'm a bit slow when it comes to Comfy and Swarm stuff.
@BattleRabbitAIart apologies but I haven't used gradio-based front-ends for almost 2 years now, so I'm at a bit of a loss when it comes to this. That being said, latest preview3-based anima models should handle upscaling much better without the need for a model swap. Maybe try RDBT-anima?