Compared to version 1, version 2 uses about twice the training data, resulting in more details and fault tolerance in the generated images. Additionally, the size of Lora is only a staggering 18MB, with no compromise in image quality after testing.
the V1.0 is here:https://www.shakker.ai/modelinfo/b171276ec42744a182ebfde17fe6dc3f/Lines-Watercolor-Decorative-Painting-Flux-Version?mine=1
Under the same prompts, the output will feature smoother lines and more reasonable compositions. The issue of color overflow has been resolved (although some people do enjoy this effect). The color output is slightly brighter compared to version 1.
Usage remains the same
Trigger words: line watercolor
Cluttered lines can be achieved with parameters like (cluttered lines:1.5), and abstract lines with pale colors can be set with (abstract lines:1.2), etc.
Example of prompt word usage: line watercolor, The train, crossing the viaduct in the valley
, abstract background, Signed 'lionwill', watercolors, (cluttered lines:1.5), pale colors, (abstract lines:1.2),
Marked sections can be replaced with any desired prompt words and signature for the same effect!
When used with other Loras, other weights can be appropriately reduced, or using equal weights will still generate images.
Comfyui is recommended for running as it is less likely to overflow memory. Looking forward to the generated images!!!
Weight test: