A proof of concept for Flux Lora training.Training Set: 20 comic book posters.Training Steps: 1000.Use the trigger word "comic style" or "comic style poster" (important).Recommended strength: 1.0 .Known Issues: hands are sometimes drawn bad.

Comic Style

comicstyle-000050.safetensors

v0.1

Chrono's Comic StyleImportant info for prompting:Trigger is:"ckcmc comic style"additional useful tokens would be:chiaroscuro, inkstyleThis is my first attempt on a comic style so please don't punish me and I hope you like it xDIt's not 100% perfect I think because I always tend to get the same-ish faces, but I really dig the style tho, the LoRA was trained on one image I gen'd "accidently" but liked it so much, that I decided to make a LoRA out of it.It's the Lich with the red hair.All the prompts I used are included in the example images to get you going!Trained on: <a rel="ugc noopener noreferrer nofollow" href="http://Flux.dev" target="_blank">Flux.dev</a>Images made with: <a rel="ugc noopener noreferrer nofollow" href="http://Flux.dev" target="_blank">Flux.dev</a> fp8Sampling method: EulerSampling steps: 20-30Distilled CFG Scale: 3.5 but you can experiment as wellClipskip: 1I am using Hires.fix with the following settings:Upscaler: 4x_NMKD-Superscale-SP_178000_GHires steps: 10Denoise: 0.20Upscale by: 1.2x - 4.0x

Files with SHA256: 8432b2e5a2ce16cbaf017f5ae330ebaa0982a2dd609be121899381df6f7b3a48