Anyline + MistoLine | High-Quality Sketch to Image

In this workflow, you can utilize MistoLine model to transform your sketches into high-quality, detailed images. By leveraging advanced deep learning techniques, MistoLine ensures that every line and detail of your sketch is accurately captured and enhanced. With the addition of ControlNet, MistoLine provides even greater precision and customization, allowing you to generate artwork that meets your exact specifications.

ComfyUI Workflow

ComfyUI MistoLine Workflow
Want to run this workflow?
  • Fully operational workflows
  • No missing nodes or models
  • No manual setups required
  • Features stunning visuals




1. What is MistoLine?

MistoLine is an advanced deep learning model designed to generate detailed and aesthetically pleasing line art from input images. Utilizing the SDXL-ControlNet framework, MistoLine is trained on a vast and diverse dataset of artworks, allowing it to extract and reproduce the essential lines and edges that define an image's structure and form. It demonstrates high accuracy and stability, adapting to various types of line art inputs, including hand-drawn sketches and model-generated outlines. Capable of producing high-quality images with a minimum resolution of 1024px on the shorter side, MistoLine excels in generalizing across diverse line art conditions, eliminating the need for multiple ControlNet models for different preprocessors. This makes it a perfect tool for "Sketch to Image" transformations.

MistoLine examples

The key features of MistoLine include:

  1. High-quality lineart generation: MistoLine produces clean, crisp, and well-defined lines that capture the essence of the input image.
  2. Versatility: The model can handle a wide range of art styles, from realistic to stylized, making it suitable for various artistic projects, including "Sketch to Image" conversions.
  3. Customization options: MistoLine offers adjustable parameters to fine-tune the lineart output according to your specific needs and preferences.
  4. Efficiency: With MistoLine, you can generate lineart quickly and effortlessly, saving valuable time in your artistic process.

2. How MistoLine Works

MistoLine is powered by a convolutional neural network (CNN) that has been trained to translate input images into high-quality lineart. Let's break down the process.

2.1. The Model's Structure

MistoLine uses an encoder-decoder structure with skip connections. The encoder takes your input image and breaks it down into smaller, more manageable pieces, capturing important features like edges and shapes. The decoder then puts these pieces back together, creating a detailed lineart output. The skip connections help preserve fine details throughout the process, ensuring an accurate "Sketch to Image" translation.

2.2. Training the Model

To train MistoLine, the developers used a huge dataset of original artworks and their corresponding lineart. The model learned by comparing its generated lineart with the ground truth lineart, and it kept adjusting itself until it could produce results that were very close to the real thing.

The training process used a combination of techniques to ensure the generated lineart was both accurate and visually appealing. This involved a lot of complex math and computational power, but the end result is a model that can create stunning lineart with ease, facilitating efficient "Sketch to Image" transformations.

2.3. Inference

Once trained, MistoLine can be used for inference, where it takes an input image and generates the corresponding lineart. The inference process is relatively fast, allowing for quick generation of lineart from new images, perfect for "Sketch to Image" applications.

For more information, please visit github.

This workflow uses the MistoLine-SDXL-ControlNet developed by

Want More ComfyUI Workflows?