Wan 14B

(Trained on still images)

This LoRA loves to generate at larger resolutions. 1408x1856 is the largest I can go without getting funky mutations, and at that res the details really shine. It definitely needs the trigger words to keep the style on the rails. Also, I've had fantastic results by doing a 2x latent upscale and then sampling several more steps. The examples in the gallery that look much more detailed have been created this way. If you're a Comfy user, most images have workflows embedded - feel free to explore the settings. I hope you have fun with this one!

Flux-1.Dev

As opposed to my Detailed Vector LoRA, this LoRA is intended to create more simple vector illustrations (fewer shapes). Flux sometimes does a great job with vector graphics, but if you include other strong keywords (e.g. an art style or something famous/well known) sometimes it just won't do it, or will partially do it. Standard trigger is "vector illustration", but you can force clipart level simplicity by also including "simple", etc. Here are some examples where the LoRA helped. Sometimes it helps a lot, sometimes just a little.

"Vector illustration, tiny dinosaur roars in a toy box, scaring away the other toys. He feels lonely and bored. He sees a stuffed teddy bear and decides to attack it. He bites its ear and pulls its fur."

"Vector illustration of a city skyline at night"

"Vector illustration of streets and buildings full of life inside a light bulb, neon lights"

Please post your creations! I love seeing what you all make :D