Anima HaDeS 1.0

Hey everyone,

I’ve been experimenting with Anima lately and decided to put together a dedicated fine-tune to see how far this 2B architecture could really go. It’s my first release in a while, and honestly, the results surprised me enough that I wanted to share it with the community.

Recommended Settings

• Steps: 30 – 50

• CFG: 4.0 – 5.0

• Resolution: 1024×1024 or standard 1MP portrait/landscape ratios

• Sampler: er sde

You don't need heavy quality tag spam to make it work. Just describe the scene, character, and style naturally or with standard tags.

Feel free to try it out, push it with some challenging prompts, and drop your feedback or generations below. Hope you enjoy using it as much as I enjoyed tweaking it!

Important Note: Required Files

Please note that this release only includes the UNet weights. To generate images with it, you will need to pair it with the standard base components:

• Text Encoder: Make sure to load the base Qwen text encoder used by Anima.

• VAE: Use the standard Anima VAE.

In ComfyUI or your preferred loader, simply load this model into your UNet/Diffusion Model slot and hook up the base Text Encoder and VAE as usual.

HaDeS