This controlnet is a scribble model in the SDXL series trained with an anime style. It specializes in completing sketches and regenerating drawings or photos while maintaining their shapes.
The recommended preprocessor to apply to images input into controlnet is lineart realistic.
normalとhardとveryhardのバージョンがあります。
このcontrolnetはアニメ調の学習をしたSDXL系列のscribbleモデルです。 落書きを描き上げたり、絵や写真の形状を維持しながら再生成するのに特化しています。
controlnetに入力する画像に適用するプリプロセッサはlineart realisticがおすすめです。
プロンプトは正確に入れたほうが良いです。
Description
FAQ
Comments (21)
normalとhardとveryhardの使い分けは何でしょうか?
its working very well for me! Thank you for sharing!!!
最高です。openposeでどうにもできなかったものがこちらなら実現できます。本当にありがとうございます!
Can't figure out what I'm doing wrong. I'm using Forge. I set ControlNet to Lineart, select lineart_realistic, this as a model. Run the preprocessor. Press Generate and the final result is not even close to my sketch
Interestingly enough sometimes I see the outlines of the result, but it goes over the generated image that's not close to the sketch. Also is it supposed to be for txt2img or img2img? I'm using with txt2img
I didn't try this specific model yet, but previously I had good lineart results by using a lowish weight and ending early.
@chernobylguy It works reasonably well for me in t2i on Forge. Have you checked the preprocessor output and confirmed it looks as you'd expect? Also, check that there aren't any extra preprocessor images being generated
@Machi yeah, preprocessor output looks as expected. I have a sketch with white bg and black lines, preprocessor gives me basically the same thing but inverted. I only have the result and this one preprocessor image in the output after generation. What other settings do you use in the ControlNet unit?
@chernobylguy I've been doing something like this https://files.catbox.moe/ra51i5.png with earlier or later ending steps depending on how much I want to preserve the lines
@Machi this is so weird. I copy your settings, the result is the same, generation just ignores the controlnet
I've tried my own old "sketch". https://files.catbox.moe/1emgv2.png
I've tried some meme from the internet https://files.catbox.moe/xroj38.png
Both are not even close to the reference. Am I doing something wrong?
@chernobylguy Are you using an XL format model as your base model?
@hakoniwa I'm trying to use PonyAutismMix, Confetti version here, I believe it is the XL model
maybe add the ComfyUi Workflows for the Examples images
Tell me, to enter a realistic photo, to get it, but in the style of anime, do I need to use scribble, or something else? Because I don't get the desired result
me neither
こんにちは。A1111のWebUIだとLineartをTypeとして選択しても出てこないのですが、どうしたらいいですか?
Error: Could not load the controlnet model! Reason: Can't find a yaml config file for the ControlNet! Please download the YAML config file (from where you downloaded the model), and put it next to the model. Please use the same name as the model file, e.g. 'foo.yaml' if the model is 'foo.safetensors'
possibility of a fp16 version?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.

