The model uses a large SDXL1.0 base model for training, similar to Midjourney's 3D niji style. I defined it as the Q version blind box style. It is mainly used for generation, characters, animals, and can also be used to generate other items

Reminder: You can use SD1.5's traditional hint writing method, that is, use TAG, or use natural language to write prompts (similar to midjourney). I recommend natural language.

Example hint: “A cute boy is playing soccer, wearing a red and white soccer jersey with shorts and soccer shoes, with a soccer field in the background”

After translation: A cute boy playing soccer, wearing a red and white soccer jersey with matching shorts and soccer cleats, a soccer field in the background

Resolution: 768x1024, 768x1152, 928x1232, 960x1280 (recommended), 1024x1024 (these are a few resolutions I usually use, mainly 2:3 and 4:3. You can try it yourself, the recommended resolution range is 768-1536)

(Online raw images, up to 1024, so online raw images are recommended 768x1024, local suggestions 960x1280)

Sampler: Euler a or DPM++ 2m karras

Number of steps: about 30 steps Clip: 2

Trigger words: owo style, chibi

positive hint words: 8K, Official art

reverse hint words: easyNegative, badhandv4, (worst quality, low quality: 1.3), logo, watermark, signature, text, Eu