Reddit上有人尝试把阿汤哥的壮志凌云片段用AI转成漫画，使用的ControlNet （github.com/lllyasviel/ControlNet）

作者分享了他制作的经验：

In my first test (the old version of controlnet) I wanted to do an anime style but it turned out to be a combination between anime and American cartoon, so with controlnet 1.1 I did this test to see if I get rid of the references to "scanner darkly" lol and what It looks more like a real anime. I only have one day testing controlnet 1.1 and I think it still has a lot to offer and more combinations to try! lots of potential!
在我的第一个测试中（旧版本的controlnet），我想做一个动漫风格，但结果是动漫和美国卡通的结合，所以在controlnet 1.1中我做了这个测试，看看我是否摆脱了 "黑暗扫描仪 "的引用，以及它看起来更像一个真正的动漫。我只有一天时间测试controlnet 1.1，我认为它仍然有很多东西可以提供，还有更多的组合可以尝试！有很多潜力

I think the new 1.1 models are so powerful that "img2img alternative script" is no longer necessary but maybe it depends on the situation. Openpose full is very powerful and keeps everything in place in the characters! along with depth mida (what an incredible training job these guys did) I tried a lot with "img2img AS" but it doesn't let me change the style much so I decided to start testing without it! and the fixed seed (although sometimes it depends on the noise, the seed passes irrelevant for some reason).
我认为新的1.1模型非常强大，"img2img替代脚本 "已经没有必要了，但也许要看情况而定。Openpose full是非常强大的，它能让所有的东西在人物中保持原位！随着深度mida（这些人做了多么令人难以置信的训练工作）我尝试了很多 "img2img AS"，但它不能让我改变很多风格，所以我决定开始测试没有它！还有固定的种子（虽然有时取决于噪音，种子通过不相关的一些原因）。

🔗 www.reddit.com/r/StableDiffusion/comments/12rvhef/top_gun_anime_v2_controlnet_10_vs_11/ http://t.cn/A6NxZAZp