强化写实的同时赋予无拘无束的创意空间，实现各类生动场景的即时创作，但也容易手崩。推荐负面:(worst quality, low quality, illustration, 3d, 2d, painting, cartoons,sketch),tooth,openmouth,negativeXL_D,unaestheticXLv13,训练用触发词：Duobian采样步数：20-150，分辨率512*768起步CFG：写实建议4.5以下，幻想类7左右采样方法：eulera或者 3SDE karras

SDXL_实用大底模

既能写实，又能搞创意，实用于各类场景，但也容易手崩.
推荐负面:(worst quality, low quality, illustration, 3d, 2d,painting,cartoons,sketch),tooth,openmouth,negativeXL_D,unaestheticXLv13,
训练用触发词：Duobian，
采样步数：20-150，
分辨率512*768起步CFG：写实建议4.5以下，幻想类7左右采样方法：eulera或者3SDEkarras
开AD

SDXL_实用多变大底模_V1.0.safetensors

dgirlCN_sdxl_ever-changing big model

While enhancing realism, it also gives an unfettered creative space, realizes instant creation of all kinds of vivid scenes, breaks creative boundaries, and allows imagination to flourish. Thank you Cayden, Meow, Wan Xiang, YINGPING, Brother Laoyao, and Mo Bai for their help in the test, great love! Those who are interested in local testing can join the group 651194700 , work hard @ management (resistant to manufacturing), won't use it, you can leave a message ha Recommended negative: see reference chart (Moved from Zhan Ye Xiao) Easy to use! Trigger words for training: Daote, Number of sampling steps: 20-150, The resolution starts at 512 x 768 CFG: Realistic recommendation 4.5 or less, fantasy category about 7 Sampling method: euler a or 3SDE karras Flexible adjustment of the number and amplitude of redrawing steps Still testing, there is a problem with the raw image

Files with SHA256: bda276d077431686218e0c4d9d19e3faad8582876464d04bc9982ddc8fd9018a