Zabadu's MIX

This is the first time I managed to make a full checkpoint, though it is a merge checkpoint, not a personally trained one.

Update: I apparently uploaded the wrong version for the Realism v2 model, it was the same file as v1. I have uploaded the proper version now and replaced the last one. If you already downloaded the previous version of Realism v2, please replace it with this new one.

Note: If anyone is confused about the v1 models, these are the same files as before (they were originally named v1 for the Drawing, v2 for the Anime, and v3 for the Realism). I only changed the names to match the new v1/v2 structure as I wasn't originally intending to do new versions of these models. So, you can either keep the originals, or change their names to match the new names, but you don't need to redownload the v1 files unless you didn't have them already.

V1

(These were renamed, but they are still the same files)

This is mostly a mix of Boytaku and Wai v15, with a little bit of CyberIllust (to add some realistic capabilities), and some art style Loras I've made, including my MaleMix v2 and Gremlin models, as well as a few other styles I haven't posted anywhere.

For Detailed Drawing v1 the overall style is a detailed colored pencil drawing style, which can get more realistic or more anime/cartoon with prompting, but would likely need other style loras to really push it further.

For Cute Anime/Toon v1 I merged Detailed Drawing v1 with Wai v16, and some loras I've made to push a more anime/toon style, with a slightly cuter overall aesthetic.

Realism v1 Is my attempt to push more realism over artistic styles. This is another merge, but this time I used some of the Cute Anime/Toon v1, with more CyberIllust for realism, and a realistic/photo lora I trained specifically for this merge. I also included some plastic anime/blindbox figure loras, because this was my own personal goal for using this new merge.
So adding "photo, realistic" should be enough to make the realistic images, and "plastic, pvc figure" for more anime style figures and "plastic, blindbox, cute" for the smaller chibi-ish figures.

NOTE: Because of the inclusion of cuter figurine model, and other cute styles, you may need to include "mature" in your positive, or words like "cute" in the negatives, otherwise you might get more teen-ish looking outputs.

V2

These versions used a new merge, though the main components were mostly the same I just adjusted the various strengths and worked from a singular base merge for all three models, which included Boytaku and Wai v16, but used the Retro Vintage Realism and Nova Kemono, for some base photorealism and furry/anthro art. It also included my MaleMix v2 and Gremlins loras, as well as a few other art style loras (at a lower strength than the previous versions).

For Detailed Drawing v2 I included Cyber Semi-real, and additional traditional/detailed art style loras to help capture a different detailed hand-drawn style from the previous version. Also with v1 I noticed the style wouldn't blend well with certain loras, usually overpowering them, this new version seems to handle style loras better.
For Cute Anime/Toon v2 I pushed it more toward the western toon side. This one included Cyber Anime, and more western cartoon and comic based art style loras. It still focuses on the cuter aesthetics, and smaller body proportions. Since the previous Anime/Toon model was more mixed, there are still some similarities.

For Realism v2 I included Cyber Realistic, instead of using it in the base model, and I also added a few photo-real style loras, mostly aimed at more vintage/analog film style photos. I didn't include as much for the anime figures this time, so it should give less cutesy outputs, but may need additional loras if you are looking to make realistic figurine images.

The sample images only use this checkpoint merge, and no additional loras. They were however upscaled, with inpainting to fix hands and faces.

Detailed Drawing v1

Detailed Drawing v2

Cute Anime/Toon V1

Cute Anime/Toon v2

Realism V1

Realism v2

I don't have much to say, as they seem to handle the things I've made with them so far fairly well, and I haven't noticed any particular issues.

But feel free to drop a comment if you run into anything.

All generations are made using Fooocus, and upscaled to 2048x2048.

All of my models may be used for your own training and/or merges.

Search AI Models & Addons

very few results