What Kohya SS is and why you'd use it

Kohya SS is a graphical interface built on top of the Stable Diffusion machine learning model. Instead of typing commands into a terminal, you get buttons and text fields — it's designed to make training custom AI image models less intimidating for people without programming experience.

The tool lets you take a folder of images and train Stable Diffusion to recognize specific people, objects, or styles. Once trained, you can generate new images in that style or of that subject. Kohya SS handles the technical setup that would otherwise require writing code or managing Python packages yourself.

The name comes from its creator, a developer who goes by Kohya. "SS" stands for Stable Diffusion. You'll see it called Kohya SS, Kohya, or sometimes the Kohya GUI interchangeably.

Key Takeaways

  • Kohya SS requires Python 3.10 or 3.11 installed on your computer first, plus a graphics card (GPU) that supports CUDA if you're on Windows or an Apple Silicon Mac if you're on macOS.
  • You download the Kohya SS code from GitHub, then run a setup script that installs the dependencies it needs to function.
  • The interface launches in your web browser even though the software runs locally on your machine, not on a remote server.
  • Training a custom model requires a dataset of images (typically 20 to 100 images), which you organize into folders before starting the training process.
  • The entire setup and first training run can take 30 minutes to several hours depending on your hardware and how many images you're training on.

System requirements before you start

Kohya SS runs on Windows, macOS, and Linux. The hardware you need depends on your operating system and what GPU you have access to.

On Windows: You need an NVIDIA graphics card that supports CUDA (most NVIDIA cards from the last five years do). If you have an older NVIDIA card or an AMD card, the software will still run but much more slowly. You also need Python 3.10 or 3.11 — not 3.12 or later, because the dependencies haven't caught up yet.

On macOS: If you have an Apple Silicon Mac (M1, M2, M3 chip), Kohya SS will use your GPU automatically. If you have an Intel Mac, it will run on CPU only, which is slow for training. You need Python 3.10 or 3.11.

On Linux: You need CUDA support (NVIDIA GPU) or ROCm support (AMD GPU). The setup is more involved than Windows or macOS because you may need to install CUDA drivers separately.

Regardless of your OS, you need at least 8 GB of RAM and 10 GB of free disk space. Training runs faster with 16 GB of RAM and a dedicated GPU, but it's not required.

Installing Python and checking your version

Open a terminal (Command Prompt on Windows, Terminal on macOS or Linux) and type python --version. If you see Python 3.10.x or 3.11.x, you're ready to move forward. If you see 3.12 or later, or if Python is not found, you need to install or downgrade.

Download Python 3.11 from python.org. On Windows, during installation, check the box that says "Add Python to PATH" — this lets you run Python from the terminal. On macOS, the installer handles this automatically.

After installation, close your terminal completely and open a new one. Type python --version again to confirm the new version is active.

Downloading Kohya SS from GitHub

Go to github.com/bmaltais/kohya_ss in your web browser. On the right side of the page, you'll see a green button labeled "Code". Click it, then click "Download ZIP".

Extract the ZIP file to a folder on your computer. The location doesn't matter much, but avoid paths with spaces or special characters — use something like C:\kohya_ss on Windows or ~/kohya_ss on macOS.

Open a terminal and navigate to that folder. On Windows, type cd C:\kohya_ss. On macOS or Linux, type cd ~/kohya_ss. You should now be inside the Kohya SS folder.

Running the setup script

Inside the Kohya SS folder, you'll see several files. Look for setup.bat (Windows) or setup.sh (macOS and Linux).

On Windows, double-click setup.bat. A terminal window will open and begin installing dependencies. This can take 5 to 15 minutes depending on your internet speed and whether you have a GPU. You'll see text scrolling past — this is normal. Let it finish without closing the window.

On macOS or Linux, open a terminal in the Kohya SS folder and type bash setup.sh. The same process happens in the terminal window you're already using.

When the setup finishes, you'll see a message or the terminal will return to a prompt. Do not close the window yet.

Launching the Kohya SS interface

After setup completes, look for gui.bat (Windows) or gui.sh (macOS and Linux) in the same folder.

On Windows, double-click gui.bat. On macOS or Linux, type bash gui.sh in the terminal. A new terminal window will open with startup messages. You'll see a line that says something like "Running on http://127.0.0.1:7860" — that's the address where the interface is running.

Open your web browser and go to http://127.0.0.1:7860 (or localhost:7860). The Kohya SS interface will load. It looks like a web page, but it's running on your own computer — nothing is being sent to a server.

Keep the terminal window open while you use Kohya SS. Closing it will shut down the interface.

Preparing your training images before you start

Before you train a model, gather the images you want to train on. For a person, you typically need 20 to 100 photos from different angles and lighting. For an object or style, 20 to 50 images usually works.

Create a folder on your computer — for example, my_training_images — and put all your images in it. The images should be JPG or PNG files. Kohya SS will resize them automatically, so they don't all need to be the same size.

In the Kohya SS interface, you'll point to this folder when you set up a training job. The interface will read the images from that location and use them to train the model.

Frequently Asked Questions

Do I need a GPU to use Kohya SS?

No, but training will be very slow without one. On CPU only, training 100 images can take 8 to 12 hours. With a modern GPU, the same job takes 30 minutes to 2 hours. If you don't have a GPU, you can still learn how the interface works, but you may want to use a cloud service like Google Colab instead.

What do I do if the setup script fails?

The most common cause is Python version mismatch. Open a terminal in the Kohya SS folder and type python --version to confirm you have 3.10 or 3.11. If you have 3.12 or later, uninstall it and install 3.11 instead. If setup still fails, check that you have at least 10 GB of free disk space.

Can I close the terminal while I'm using Kohya SS?

No. The terminal window is running the server that powers the interface. Closing it will disconnect you from Kohya SS. You can minimize it, but leave it open while you work.

How do I stop Kohya SS when I'm done?

Close the terminal window where you ran gui.bat or gui.sh. You can also press Ctrl+C in the terminal to stop the server gracefully. The web browser tab will no longer load the interface after that.

What if I see an error about CUDA or GPU support?

On Windows with an NVIDIA GPU, this usually means CUDA drivers aren't installed. Download the CUDA Toolkit from NVIDIA's website and install it, then run the Kohya SS setup script again. On macOS or Linux, check your GPU documentation for driver installation steps specific to your hardware.