The LoCon/LyCORIS system is designed for SD2.1 series, targeting WD1.5beta3. It aims to produce outputs reminiscent of black and white anime line drawings.
Please install the extension feature of LoCon. Better results can be achieved by using Hires.fix with a 1.5 scale-up at 512x768 resolution.
For positive inputs, we recommend "monochrome, lineart, <lyco:sd2ly_line:0.65>".
For negative inputs, we suggest "(mayng:1.4), fullcolor, photo, shadow".
Please refer to the prompts from the sample images.
mayng(SDv2 Negative Embedding)
Description
FAQ
Comments (8)
Unable to run
RuntimeError: The size of tensor a (768) must match the size of tensor b (1024) at non-singleton dimension 1
@kirada I think you're using it with a 1.5 model... It's a 2.1 model Lora, working fine with me !
@kirada that's a lycoris that was made for stable diffusion 2.1, so it's normal to getting error on an SD1.5 models
This is how I got this LyCORIS to work:
+ Installed the LyCORIS webui extension from https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris. (A new LyCORIS tab appears in extra networks)
+ downloaded the sd2ly_line.safetensors and saved it to ..\stable-diffusion-webui\models\LyCORIS
+ downloaded and used a hakoMay model from https://huggingface.co/852wa/hakoMay/tree/main (I used hakoMayB)
+ I got an error "# A tensor with all NaNs was produced in Unet", To correct it, in a1111 webui go to Settings > Stable Diffussion > [check] Upcast cross attention layer to float32
What version is your stable-diffusion? 1.5 or 2.0？
@guderianw312wu I'm not using a raw stable diffusion release, i used a hakoMay model from https://huggingface.co/852wa/hakoMay/tree/main
The same problem occurred when I used comfyui's clip loader node to import mayng. It turns out that mayng is also a Lora model (the weight file is too small), so it says you need two Lora loaders, one connecting model and positive tip words, and mayng's lora connecting negative tip words.