试着训练了一个极简风格的locon，感觉还不错，分享给大家看看。建议使用以下提示词格式进行出图：采样器：DPM++ 2M SED Karras迭代步数：30~50出图底模：<a rel="ugc" href="https://civarchive.com/models/136389/kohaku-xl">kohakuXL_alpha7</a>出图分辨率：1024*1024提示词格式：Minimalism \(icon\)+Black theme（颜色主题）+bird（实际物体）+best quality。提示词中还可以适当的添加其他诸如colorful、hdr等提示词增加画面色彩、光影。locon训练的还不是很充分，表现也不是特别稳定，其中1girl,1boy,male,female,child等词并没有能够达到预期表现，后续会尝试优化。因为本次测试的不是很充分，还希望大家能够多多见谅，本页最后为测试图，供大家参考，希望对大家有帮助。——————————————————————————————————Tried to train a minimalist locon and felt good enough to share.It is recommended to use the following cue word format for the locon:Sampler: DPM++ 2M SED KarrasIteration steps: 30~50Map base model: <a rel="ugc" href="https://civarchive.com/models/136389/kohaku-xl">kohakuXL_alpha7</a>Map resolution: 1024*1024Cue word format: Minimalism \(icon\) + Black theme (color theme) + bird (actual object) + best quality.Cue word can also be appropriate to add other words such as colorful, hdr and other cues to increase the screen color, light and shadow.Locon is not fully trained, and its performance is not particularly stable. 1girl, 1boy, male, female, child and other words do not meet the expected performance, and we will try to optimize it later.Because this test is not very adequate, I hope you can forgive us, the last of this page for the test chart for your reference, I hope to be helpful to you.测试图（Test chart）：<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0a5b4969-21c1-484b-8bce-c5e5e24c3a11/width=525/0a5b4969-21c1-484b-8bce-c5e5e24c3a11.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/8ac5c74c-d1f2-41e0-a894-c3d5176081c5/width=525/8ac5c74c-d1f2-41e0-a894-c3d5176081c5.jpeg" />

Files with SHA256: 7760b60570c8397527a08de2bd5288f83c310ad4edc686c156b8a41845c6d76f