Hacker News new | past | comments | ask | show | jobs | submit login

I remember I tried to get an early version of Midjourney to draw something in ligne claire-style, no matter how I expressed it, it failed miserably at that. It was an early "Yeah, these models are actually pretty overfit" moment for me.

I haven't tried lately, though. I would assume it's better now.




Flux-1-Schnell does pretty decently just by prefixing a prompt describing the subject without style information with “An illustration in ligne claire style.”


Fine-tuning stable diffusion tend to work pretty well for this kind of things.


Stable Diffusion 3.5 Large does pretty well out of the box (similarly to what I noted for Schnell in another comment, just using “An illustration in ligne claire style…” as the start of the prompt) without any custom fine tuning.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: