If you like my work and want to support me, or are interested in information about my upcoming projects or updates to my models, visit: https://ko-fi.com/devildonia.

Ultimate Anime from HaDeS 2.5D v16 HQ

We have reached the zenith of this base model. The quality jump is huge in this update and the typical configurations of previous versions need a deep touch-up. This v16 already focuses completely on 2.5D. For 2D images I will continue developing and improving my other model: Animai (which has just been updated to version 2.1).

Recommended settings:

Sampling method: DPM++ 2M SDE Karras / DPM++ 2M Karras

Sampling Steps: +30

Hires.fix Upscaler: 4x-UltraSharp

Hires Steps: half of the Sampling Steps (example: if the sampling steps are 30, the hires steps must be 15).

Denoising strength: between 0.35 and 0.45

Upscale by: It will depend on your PC, I personally use 2

CFG Scale: between 7.5 and 10.5 depending on the desired detail. I personally use 10.5

With this configuration you will achieve very high quality results, but obviously I am sure that by fiddling around a little you will improve even more. I can't wait to see your creations and how you fill this wall with your images.

I hope you enjoy this update and give it lots of love in the form of likes and reviews. See you in the next update before the jump from Anime from HaDes to Stable Diffusion XL which is already in development.

Update v14 HQ

Anime from HaDeS now consists of two independent models that are mixed with great care to achieve that effect of mixing two and three dimensions, preserving a cartoon or anime look.

This update adds more depth and explores the limits of Stable Diffusion 1.5's ability to create 2.5D characters while retaining realistic 3D backgrounds and environments.

Update v12 HQ

A big leap in quality in this update. For all those who enjoy manga and anime, this checkpoint recreates the styles of many mangakas from the land of the rising sun. Test it and share your creations to help me improve it more and more.

Update v6 ReDuX

A style more inclined towards animation, without losing the essence of keeping a cartoon realism. A very fun base model, which works great above all with almost any LoRa. The only limit is your imagination.

V4

This model perfectly mixes realistic images with 3D anime. It recreates various styles, poses, characters, animals and is designed to be a complete model covering: Fantasy, realism and imaginary worlds.

I have been developing this model for a long time, it is currently in version 4 and I decided to upload it so that everyone can enjoy it, test it, share their impressions to continue improving it.

Here are some configurations that work quite well with this checkpoint:

Sampling method: DPM++ 2M Karras

Sampling steps: 25-35

Clip skip: 2

Upscaler: 4x-Ultrasharp

Hires Steps: 10-20 (15-16 gives the best result)

Denoising strength: 0.4-0.5

Cfg scale: 4-8