The goal of this lora was initially to give a finer screencap style similar to the <a rel="ugc" href="https://civarchive.com/models/273116/fine-anime-screencap-style-or-lycoris">SD 1.5</a> and <a rel="ugc" href="https://civarchive.com/models/345962?modelVersionId=502952">PonyXl</a> loras, but after several training attempts I realized that with animagine this was not possible since the model is trained with a very marked screencap style. But during the tests I noticed that it improves the color, the blacks, the saturation, the blur it generates, and also improves the generated backgrounds here is an example:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/8522b56b-337d-4436-9df1-9a24dd142cb8/width=525/8522b56b-337d-4436-9df1-9a24dd142cb8.jpeg" /> If you don't add "anime screencap" and just add the trigger word, you will also get an improvement but not the style, here is an example:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1492213b-fc6a-4387-a866-fa6817fcb178/width=525/1492213b-fc6a-4387-a866-fa6817fcb178.jpeg" />The improvement effect will depend on the character you want to generate as well as with character loras, the recommended is to use 1 weight but you can increase it to 1.5.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1bf930ff-bb5d-4a00-95c8-b41511655e93/width=525/1bf930ff-bb5d-4a00-95c8-b41511655e93.jpeg" />Also I recommend using the samplers: euler a or eulerThat's all enjoy it!

Files with SHA256: d30ce8245463cfdad9fea06af216fd5c30d93205b1b130983fa64a6f9751cb74