/vt/ - Virtual Youtubers

https://huggingface.co/kohya-ss/controlnet-lllite
examples are all of 1.00 strength, but you just lower the strength you can still catch the idea of the image and have it be more flexible, while still remembering key details, assuming it knows what its looking at, the prompt can help as well. you can also change the end step of the CN and get other results, you just have to balance but you can see this slightly higher strength has even included the text, and end step