This Flux Lora was trained to generate images in the style of vintage pulp fantasy comics "by" Frank Frazetta (I found that they are not really drawn by him, he just put his name on them, but now it's too late to change the trigger word ^^°).

Many thanks to SubtleOne who gave me access to his training data for the model!

As the training data was dark fantasy, the model prefers medieval fantasy style images but it can also generate any other setting.

The style is pulp fantasy vintage comic with high contrast and simple line arts.

Training was done for ~8000 steps with SimpleTuner. Lora rank was 6 - I found lower rank works better for style training, as the model does not memoize the training data too much.

The example images are non cherry picked. Prompts were randomly made with ChatGPT. None of the images and prompts has any connection to the training data. Add "Illustration by Frank Frazetta:" in front of the prompt (or add "Illustration by Frank Frazetta" at the end). I didn't used the trigger word "comic" to avoid that the AI is putting speech bubbles into the images.

Use the lora at full strength to get the effect, or reduce the strength to make the look less dark-fantasy and more bright comic.

I don't care what you do with the models, but I don't want to be spammed with ???? pictures of this model on civitai, so please don't use the model for ???? or at least don't spam this gallery with obscene pictures (and let's face it, the way those old pulp comics depict women and men isn't really sexy).