Anima
Go to the official Anima Huggingface repo for answers.
You need to use qwen_3_06b_base.safetensors for text encoder, and qwen_image_vae.safetensors for VAE.
Installing and running
The model is natively supported in ComfyUI. The above image contains a workflow; you can open it in ComfyUI or drag-and-drop to get the workflow. The model files go in their respective folders inside your model directory:
animax.safetensors goes in ComfyUI/models/diffusion_models
qwen_3_06b_base.safetensors goes in ComfyUI/models/text_encoders
qwen_image_vae.safetensors goes in ComfyUI/models/vae (this is the Qwen-Image VAE, you might already have it)
Generation settings
The preview version should be used at about 1MP resolution. E.g. 1024x1024, 896x1152, 1152x896, etc.
30-50 steps, CFG 4-5.
A variety of samplers work. Some of my favorites:
er_sde: neutral style, flat colors, sharp lines. I use this as a reasonable default.
euler_a: Softer, thinner lines. Can sometimes tend towards a 2.5D look. CFG can be pushed a bit higher than other samplers without burning the image.
dpmpp_2m_sde_gpu: similar in style to er_sde but can produce more variety and be more "creative". Depending on the prompt it can get too wild sometimes.
Description
FAQ
Comments (8)
ive been using finetuned, yes it has better quality than the based but prompt adherence just worse, i wish it will fixed soon
Do you have any examples of prompts that Im having trouble with? That would help me.
in general using natural language and simple or complex pose, compared to the anima base
like the chracter doesnt do the prompt u want
@Seii1 Right now, Im focusing on the booru-like style. In the future, Ill add descriptions in the native language to the caption. I think this will improve this aspect.
how can i make results more creative?
To get more creative / wild / interesting results, the best thing you can do right now is play around with the prompts themselves.
I trained this model on a bunch of booru-like style prompts that were themselves generated by LLMs, so it really likes descriptive tag-lists - stuff like: flat colors, sharp lines, rectangular format, modern style, stylized art, clean shading, vibrant palette, dynamic composition, surreal vibes, retro-futuristic, and so on.
Right now I’m not gonna sit here and test 150 different tags because this is still kind of an intermediate / testing version. The goal is to see whether people even like the direction, if the overall vibe feels good, if it’s worth pushing further.
By version 1.0 I’m planning to put together a much more detailed & juicy prompt guide with really strong, proven tags.
So for the moment — just experiment with adding all kinds of art/style/quality/aesthetic tags at the end (or beginning) of your prompt and see what starts to happen. Have fun!
Rityak, I need your professional opinion: is it worth switching from the Illustrious model to the Anima model?
Its definitely worth keeping an eye on, at the very least. There isnt much here yet, but I think there will be soon. The model is easy to train, and there are already many LORA models for it. Perhaps in six months to a year, it will become the new standard in anime.