DeepSeek’s ELI5 on what LoRAs are:

Imagine you have a big box of crayons that can draw almost anything you can think of. This box is like Stable Diffusion, a super-smart AI that creates pictures from your words. But sometimes, you want to draw something very specific, like a unicorn wearing a spacesuit, and your big box of crayons doesn’t have the exact colors or tools to make it look just right.

This is where LoRAs come in! LoRAs are like small packs of special crayons or stickers that you can add to your big box. They don’t replace your crayons—they just give you extra tools to draw specific things better. For example, if you want to draw that unicorn in a spacesuit, you can use a LoRA that knows all about spacesuits or unicorns, and it helps Stable Diffusion make the picture look exactly how you want.

In technical terms, LoRAs (Low-Rank Adaptations) are small, lightweight files that tweak how Stable Diffusion works. They help the AI focus on specific styles, objects, or details without changing the whole system. So, instead of needing a whole new box of crayons, you just add a little extra magic to the one you already have!

Fake MrBeast:






(the last one also uses the PogChamp LoRA)

Morshu:


(the first one also uses this Zelda CD-i LoRA)

Pepe Punch Meme:

Peter Griffin-ish:

Puddle of Tears:

Two Buttons Meme:

Note: I tweaked a few of the pics via inpaint sketch to remove errors (e.g. wrong number of fingers).

In all images, I used:

• Stable Diffusion WebUI Forge
• AutismMix_pony
• Shane Glines Style LoRA