This is the virtual character IP I personally trained, and the trigger word is completed with my personal name abbreviation.If you want to learn this process and avoid pitfalls in the implementation process, you can follow my<a target="_blank" rel="ugc" href="https://www.youtube.com/@shixiongsong"> YouTube</a> or<a target="_blank" rel="ugc" href="space.bilibili.com/219572544?spm_id_from=333.1007.0.0"> Bilibili</a> account（ID：啦啦啦的小黄瓜）, and I will update the complete process later.I replicated the entire process of <a target="_blank" rel="ugc" href="https://www.youtube.com/@mickmumpitz">Mickumpitz</a>, but I built all the workflows I used myself.Step 1：generate the basic image.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/82bb074c-fd38-4b2c-a999-ee5a367b92a5/width=525/82bb074c-fd38-4b2c-a999-ee5a367b92a5.jpeg" />Step 2: Enlarge and segment the image to generate multi expression characters<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9bff0902-2715-4cab-9da6-785b78473015/width=525/9bff0902-2715-4cab-9da6-785b78473015.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/4f3eb259-faab-4307-a326-14f859c58f5e/width=525/4f3eb259-faab-4307-a326-14f859c58f5e.jpeg" />Step 3: Adjust the dataset to the appropriate size and perform image restoration<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f88f318c-0819-4871-847f-85876cb6b1fb/width=525/f88f318c-0819-4871-847f-85876cb6b1fb.jpeg" />Step 4：Label the image<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1a3ef4ed-674e-44ac-aa14-b744abb0d901/width=525/1a3ef4ed-674e-44ac-aa14-b744abb0d901.jpeg" />Step 5：Training Lora model

Files with SHA256: 107cb84778f29d43f2e58fbcf89329a09bc61bde4a5cdb1690b26aec7e8345e6