Download
1 variant available
209
11.4K
2.7K
KREA 2 Version
You might be asking yourself, why do you need a Lora for a style that Krea 2 already does fairly well right out of the box. I too questioned if it was worth my time. But I also couldn't think of a better artist to celebrate 250 years of USA independence. I was also pleasantly surprised with the results when compared to what's already in the model.
I compared same seeds and prompts and only changed "A Norman Rockwell style painting" for the no Lora version and for the lora, I started with "a n0rm4n style painting" to try to exhibit two completely different triggers.
First noticeable difference was way fewer watermarks when using the lora. I manually removed all signatures for the dataset to try to reduce this. It works about 95% of the time which isn't bad when you don't have negative prompt.
Second noticeable difference was Krea 2 will try to force your image to fit into a 1950's theme even if you're trying something not related to the style or genre. When trying to use characters that Krea 2 already knows but doesn't associate with Norman Rockwell, it will try to force those characters and background into a mid 20th century style. But if you just want the paint style, but more modern topics and subjects, that's where this Lora comes in really handy.
Norman Rockwell was an American painter and illustrator celebrated for his vivid, heartfelt portrayals of American life. Born in 1894, he became famous for capturing everyday scenes that resonated deeply with the public. His work often appeared on the covers of The Saturday Evening Post, where he depicted moments of humor, innocence, and patriotism, creating images that became emblematic of mid-20th-century American culture. Rockwell's art combined realism with idealism, offering a nostalgic look at small-town America. Among his best-known works are "The Four Freedoms" series and "The Problem We All Live With," the latter addressing social issues and marking a shift to more poignant themes later in his career.
This LoRa was trained on the flux-dev-de-distill-fp8 model