Flash Manga EXPRESS
The style is loose, its expressive, and its not afraid to get things a bit wrong. So, the perfect option for this was illustrious.
The characters made from this model should come with a sketched outline around them, there are likely to be a few errant lines and the colors may come across slightly muted (this allows nice pop color effects) as if drawn on old paper. The images should feel almost childish in their drawings an anatomy while retaining enough knowledge to still make sense.
This LoRA was trained on 172 images for 10 epochs at 1024,1024 resolution. It favors working in Square, 3:4, or 4:3. I plan on expanding that down the line and adding more to the dataset to help fully enhance the effect it creates.
I personally prefer running the lora at full strength but feel free to adjust that. I run it in DPM2MM Karras. I like to run 20 steps with a cfg around 4.
This model is meant to work best on illustrious and will have noticeably muted effects on other checkpoints but I recommend trying it out, I'm seeing interested results when testing it on my other checkpoints that I'm probably going to build into a full release.