SYSTEM NOTICE

Auto translation by AI. Be sure, accuracy, nuances and authorial intent may not be fully reflected.
見出し画像

The Definitive Guide to AI Illustration Editing! How to Safely Install and Use IOPaint (Lama-cleaner) [AI Image Generation & Image Editing]

Introduction

Hello, I am Kimama / Easygoing.

In this article, I will introduce Lama-cleaner / IOPaint, a tool that allows you to easily edit illustrations.

An anime illustration depicting a snowy airport runway, an airplane, and a smiling woman.
Snowy Airport

Lama-cleaner / IOPaint is a free image editing tool

Lama-cleaner / IOPaint is an open-source image editing tool that anyone can use easily.

Editing with Lama-cleaner / IOPaint

A comparison image showing the before and after of removing an airplane from an illustration using IOPaint.
Removing an airplane
A comparison example showing damage (scratches) on an illustration neatly repaired using IOPaint's restoration function.
Repairing damage

With Lama-cleaner / IOPaint, you can edit illustrations in just a few seconds simply by selecting the area you want to edit.

The editing accuracy is also very high, to the point where you wouldn't notice the edits unless you knew the original illustration or observed it closely while zoomed in.

LaMa is a groundbreaking image restoration AI model

Lama-cleaner / IOPaint uses an image restoration AI model called LaMa, which was released on September 15, 2022.

LaMa: Resolution-robust Large Mask Inpainting with Fourier Convolutions

Image restoration AI models have existed for a while, such as Google's Magic Eraser, but they had drawbacks: they were computationally heavy and their accuracy dropped when the missing parts were large.

The LaMa model uses a technique called Fourier Convolutions to significantly improve computational efficiency, allowing it to run at practical speeds not only on GPUs but also on CPUs.

An anime illustration depicting a snowy airport runway, an airplane, and a smiling woman.
LaMa is fast!

Furthermore, due to improvements in computational efficiency, dynamically generating large masks during training increases training efficiency, and by referencing the entire image during restoration , it is now possible to perform accurate restoration even for large defects.

Spread rapidly with Lama-cleaner!

LaMa is a groundbreaking image restoration model, and it was released under the Apache-2.0 license, which allows for free development and commercial use.

Then, on November 15, 2022, the image correction tool Lama-cleaner, which made LaMa's features easily accessible to everyone, was also released as open source.

Lama-cleaner has a user-friendly UI, allowing you to intuitively perform image restoration.

Furthermore, Lama-cleaner runs completely locally, so unlike cloud tools such as Photoshop's Generative Fill, it is a tool that is also safe in terms of privacy.

IOPaint is a comprehensive image editing platform

Gantt chart showing the release timing of Lama-cleaner and IOPaint

Lama-cleaner was released shortly after Stable Diffusion, and it exploded in popularity as a tool for easily correcting AI illustrations, which often had flaws at the time.

Then, on January 25, 2024, it evolved into IOPaint, which integrated image restoration AI models other than LaMa, as well as features like upscaling and background removal.

Flowchart showing how multiple features were integrated into Lama-cleaner to evolve into IOPaint
Lama-cleaner evolved into IOPaint

IOPaint has become a comprehensive image editing platform that can perform inpainting and outpainting using various models.

However, in this era, other AI image generation tools were also evolving rapidly. When performing complex image corrections, dedicated tools like ComfyUI offered more detailed settings. Additionally, with the emergence of high-performance image editing AIs such as GPT-Image-1, nano-banana, and Qwen-Image-Edit, the use cases for IOPaint decreased, and IOPaint ceased development on August 13, 2025.

Anime illustration of a woman turning around with an airplane on a snow-covered airport runway.
Simple and easy to use!

Even so, for the purpose of simply correcting using the LaMa model, IOPaint remains the easiest tool to use today, so I would like to introduce how to use it to everyone.

One-click installers carry security risks

IOPaint still allows you to download a one-click installer from the official website, but

  • Development has already ended

  • Not published on open-source GitHub

  • Distributed via Google Drive (possibility of being replaced)

  • bat file

For the above reasons, security risks cannot be ruled out, and I do not recommend using this one-click installer at this time.

Anime illustration of a woman smiling inside an aircraft hangar
Let's install it manually

Therefore, this time I will introduce how to manually and safely install IOPaint on a Windows 11 PC.


Updated 2026.1.25

Mac users can find a guide on how to install IOPaint compiled by blackcat.


Python 3.10 to 3.12 environment required

First, to install IOPaint, you need a Python 3.10 to 3.12 environment, not the latest Python 3.13 or newer.

(Although official IOPaint support is up to Python 3.10, it actually works with Python 3.12 as well.)

Here, I will explain how to install Python 3.12.

Installing Python 3.12

First, access the following distribution page for the Windows Python installer.

Screenshot of the top of the official Python website's Windows download page.
Screenshot of the section on the official Python website showing the download link for Python 3.12.10.

On this page, Python download links are listed from newest to oldest, but if you scroll down, you will find the link for Python 3.12.10. Click it to download the installer.

Python installer execution screen. Highlighting the important step of checking 'Add python.exe to PATH'.

When you run the installer, this screen will appear, so check the checkbox below and run the installation.

Now you are ready to use Python 3.12.

Creating the IOPaint folder

Next, create an IOPaint folder in any location where you want to install IOPaint.

An image showing a new folder named "IOPaint" created in Windows Explorer.
An illustration of the steps to launch the command prompt by typing "cmd" in the Explorer address bar.

Then, open the IOPaint folder, click on the right side of the address bar, type "cmd", and execute it.

A black command prompt screen will launch, so enter the following command by copying and pasting it and executing it.

For Python 3.12

py -3.12 -m venv venv

For Python 3.10

py -3.10 -m venv venv

It will take a while, but once the process is finished, execute the next command.

venv\Scripts\activate.bat
A screen showing the execution of commands to create and activate a virtual environment (venv) in the command prompt.

Then, a virtual environment (venv) will be set up inside the IOPaint folder, and the screen will switch to the following.

The state where the virtual environment has started successfully and (venv) is displayed at the beginning of the command prompt.
Virtual environment started

Installing and running IOPaint

Next, execute the following command to install the IOPaint core.

pip install iopaint
A screen showing the installation of the IOPaint core by executing the pip install iopaint command.

Once the installation is finished, execute the following command to launch IOPaint!

iopaint start --inbrowser
The process of launching the program by entering the iopaint start --inbrowser command.

How to use IOPaint

Next, let's look at how to use IOPaint.

1. Select an image

The procedure for selecting an image to edit by dragging and dropping it into the IOPaint browser screen.

Once IOPaint has launched, first click in the center or drag and drop an image to input it.

2. Select the area and execute

The process of masking (selecting) the area you want to edit with a brush on the IOPaint editing screen.

Next, select the area you want to edit, and the correction will start automatically; if it does not start, press the execute button manually.

3. Download

The procedure for saving the corrected image by pressing the download button.

Finally, download the corrected image and you are done!

Creating a startup bat file

Now, let's create a startup bat file so you can easily launch IOPaint in the future.

Download the bat file that matches your environment from the links below and place it inside the IOPaint folder.

Normal startup (CPU)

@echo off

call venv\\Scripts\\activate.bat

set ARGS= --inbrowser 

echo Launching IOPaint with arguments:
echo %ARGS:^&=% 
echo.

iopaint start %ARGS%

pause

Launch with CUDA (NVIDIA GPU)

@echo off

call venv\\Scripts\\activate.bat

set ARGS= --inbrowser --device=cuda 

echo Launching IOPaint with arguments:
echo %ARGS:^&=% 
echo.

iopaint start %ARGS%

pause

Launch with MPS (Apple Silicon)

@echo off

call venv\\Scripts\\activate.bat

set ARGS= --inbrowser --device=mps

echo Launching IOPaint with arguments:
echo %ARGS:^&=% 
echo.

iopaint start %ARGS%

pause
An example of placing the 'startup batch file (.bat)' created inside the IOPaint folder.
Place inside the IOPaint folder

Now, simply double-click to run the corresponding bat file to launch IOPaint.

IOPaint Startup Options

Finally, you can check the startup options for IOPaint with the following command.

iopaint start --help
Screen showing the list of available options by running iopaint start --help in the command prompt.

Summary: Easy Correction with IOPaint!

  • LaMa is a groundbreaking image inpainting model

  • Easy correction with IOPaint

  • Install with Python 3.12

In this article, I introduced Lama-cleaner / IOPaint.

The LaMa model is an image inpainting AI model that appeared in 2022, but it still has the performance to be used actively today.

IOPaint is also an excellent open-source image correction tool, and no other correction tool that is easier to use has appeared yet.

Anime illustration of a woman smiling at the viewer with a night view of an airport in the background
Let's try using it in various ways!

Why not try combining manual correction / IOPaint / and advanced correction using ComfyUI / Stable Diffusion webUI to achieve a comfortable image generation life where you can correct illustrations exactly as you imagine?

Thank you for reading until the end!


Updated on August 9, 2026

I introduced how to use the LaMa model using custom nodes in ComfyUI.


English Article


いいなと思ったら応援しよう!