OIA Lollipop

Lollipop is an extremely excellent 2.5D large model. She can easily handle various styles, including but not limited to second dimension, illustration, anime characters, logo design, chibi characters, etc. (Some of the drawing styles can be seen in my sample images, all are original models without direct LoRA output effects).

XL version is based on my large model Butterfly training + fusion. She has very good cross-border capabilities. She is sensitive to cue words, with accurate recognition abilities in natural language, and is an excellent model for targeting and hitting accurately.

The SD1.5 version is based on a sub-model trained + fused from the large model Flyingsnow that I released on a friendly site. Lollipop perfectly inherits the strengths of the Flyingsnow model, and based on this, integrates with a weight of 0.5 the intermediate layer and the 8th-11th output layers of Ghostmix. This allows Lollipop to perfectly inherit the strengths of Flyingsnow and Ghostmix, and fine-tune the model's colors using plugins.

Anyway, I have always believed that a good large model should have versatility rather than being too biased towards a fixed style, because this can be left to LoRA to accomplish. Therefore, when conducting the final test for this model, I paid more attention to its color, composition, lines, background, characters, and the model's response to various style cues such as Anime, Photographic, Digital Art, Fantasy Art, Neon Punk, Low Poly, Line Art, 3D Model, Pixel Art, etc.

Her performance capabilities exceed my sample images, and I look forward to Lib's buddies to discover and master more abilities, and showcase your achievements in the return image area! Give a big big big bear hug to the kids who love Lollipop, love you all!