Thanks! My goal was to fuse Misty with someone. I picked Rottytops because I knew her purple shirt and green skin would go well with Misty's orange hair. Orange, purple, and green are quintessential pumpkin/Halloween colors.

This is my only zombie idea right now, but I have a few more undead designs coming!
I have a few other zombie loras I posted in the past: https://civitai.com/user/CitronLegacy/models?tag=zombie

Thank you very much! Hearing things like this are a huge motivation to make Loras!
Character combos/redesigns are one of my favorite uses of AI. There are so many cool or funny crossover/fusion ideas.
I'm working towards creating 100 combo/redesigns in this collection https://civitai.com/collections/10770010

Agree wit both comments!
Yeah Rotty is not evil she just thinks differently because she is undead.

Yes!
Short answer in 2 steps:
1. Pick two character you want to combine and prompt them to make good quality images

2. Train a lora on the best of your AI generated images

Long answer:

Same steps as above but with extra notes. You need to find two characters that are somewhat easy to prompt the details you want. For example if a character has a really complex detail it will probably be hard to generate good training images.

For MistyTops I used two Loras with this prompt to make the training images:

Sometimes you need to adjust the weights for each lora so that you get the right details. Some my generated images for MistyTops had the wrong hair style so i had to keep generating until I got enough images with the right style.

Other characters like my Lusamine x Morrigan Lora only needed one lora because Morrigan is already so well known to Illustrious

You will noticed that for both these characters I prompted with wide hips, narrow waist, because I wanted my Lora to usually give them wide hips.

.

Anyway all that said, do you have any specific questions? I can ramble for a long time about techniques, examples, or tips but this is generally the approach unless there are specific details you want to know more about?

@mystery_kcad Don't worry about asking "basic" questions. No one knows everything and everyone is always learning new things in the AI world.

I noticed your Lora is Pony. Illustrious is much easier to train loras on (once you get used to it). I'll give you some examples of datasets I've used at the end of this comment. These examples would have bad results on Pony but they work well on Illustrious.

I've seen that problem. What kind of fusion are you going for? As an example, putting Charizard's wings on a character should be easy but putting Blastoise's canons on a character might be really hard lol. Maybe describing the anatomy you want to see will help?

If you can generated images offline then you might try doing a "lazy edit" in paint and then using image 2 image to fix your image.

I suggest getting 10 to 20 images. I often make loras with 30 or less images. More is better but I make loras all the time with small datasets lol.
Here are 6 example datasets:

1. Here is the MistyTops dataset (there are 17 image and then several duplicates cropped to portrait so that the Lora learns her face really well) - https://x.com/Citron_Legacy/status/1982244131234599406/photo/1

2. Two images duplicated and cropped into 6 image - https://x.com/Citron_Legacy/status/1950695915091525657/photo/1

3. Another 2 images turned into 6 - https://civitai.com/images/53834065

4. Six images - https://x.com/Citron_Legacy/status/1959759702624989603/photo/1

5. My OC Lillie with 23 images - https://x.com/Citron_Legacy/status/1950920323294994855/photo/1

6. Dataset with 27 images - https://x.com/Citron_Legacy/status/1950915510830572021/photo/1