10 Steps to Run Your Own Private AI Image Generator - No Cloud Required

Have you ever wanted to generate AI images without worrying about your prompts being stored on some server, or running out of credits just when you need a business-suit-wearing dragon? The solution is simpler than you think: run everything locally on your own machine. With Docker Model Runner and Open WebUI, you can create a fully private, no-subscription image generation setup that feels just like using a cloud service—but all your data stays with you. In this guide, we’ll walk through 10 essential steps to get your local AI art studio up and running. From pulling models to launching a chat interface, you’ll have your own private DALL-E in no time.

1. Understand the Big Picture: Docker Model Runner + Open WebUI

Think of Docker Model Runner as the engine and Open WebUI as the dashboard. Docker Model Runner handles the heavy lifting: it downloads image generation models (like Stable Diffusion), manages the inference backend, and exposes a standard API that any compatible client can use. Open WebUI is a polished chat interface that already knows how to talk to that API—specifically the POST /v1/images/generations endpoint. Together, they create a seamless local experience: you type a prompt in the chat, and the image appears from your own computer. No internet needed after setup, no hidden fees, and your prompts never leave your machine.

2. Check Your System Requirements

Before diving in, make sure your hardware can handle it. You’ll need Docker Desktop (on macOS/Windows) or Docker Engine (on Linux). For the model, at least 8 GB of free RAM is required for a small model—more is better if you want faster results. A dedicated GPU is optional but highly recommended: NVIDIA (CUDA) or Apple Silicon (MPS) will give you a big speed boost. If you only have a CPU, you can still generate images, but expect longer wait times. Run docker model version in your terminal to confirm Docker Model Runner is installed and ready.

3. Pull an Image Generation Model

Docker Model Runner uses a compact format called DDUF (Diffusers Unified Format) to distribute models via Docker Hub, just like any other container image. To get started, open your terminal and run: docker model pull stable-diffusion. This downloads the model (about 7 GB) and stores it locally. You can verify it’s ready with docker model inspect stable-diffusion, which shows details like the model size, format, and SHA256 checksum. The DDUF file bundles everything needed—text encoder, VAE, UNet or DiT, and scheduler config—into one portable artifact. Once pulled, the model is cached and ready for action.

4. Launch Open WebUI with One Command

Here’s the magic part: Docker Model Runner includes a built-in command that automatically wires up Open WebUI against your local inference endpoint. Simply run: docker model launch openwebui. That single command downloads Open WebUI (if needed), starts it in a container, and connects it to your local model. The result is a web-based chat interface that you open in your browser, usually at http://localhost:8080. From here, you can type prompts and generate images just like you would with any commercial service—but everything runs on your own hardware.

5. Explore the Open WebUI Interface

Once Open WebUI is running, take a moment to explore its interface. It’s designed to be intuitive: you’ll see a conversation panel on the left and a main chat area where you type prompts. You can start a new conversation, adjust settings like image size or number of images, and even view generation history. The interface is fully local—no data is sent to the cloud. You can also customize it further by installing community plugins or tweaking settings. The best part? It’s completely private, so you can generate anything without worrying about content filters or moderation.

6. Understand the API Integration

Docker Model Runner exposes a 100% OpenAI-compatible API, meaning any tool that works with OpenAI’s image generation endpoint can work here. The key endpoint is POST /v1/images/generations. Open WebUI already knows how to call this endpoint, so you don’t need to write any code. But if you’re a developer, you could also use the API directly from your own scripts or applications. This compatibility makes it easy to switch from a cloud service to local without changing your workflow. Just point your API URL to http://localhost:8080 (or whatever port Open WebUI uses).

7. Generate Your First Image

Ready to see it in action? In the Open WebUI chat box, type a prompt like “a dragon wearing a business suit, photorealistic” and press Enter. The model will process your request (this may take a few seconds to a minute, depending on your hardware) and display the generated image directly in the chat. You can refine your prompt, add negative prompts, or change the number of images. Each generation is handled locally, so you can experiment as much as you want without burning credits. Your first image might not be perfect, but that’s the fun of it—you have full control.

8. Optimize Performance for Your Hardware

To get the best speed, use a compatible GPU. On NVIDIA, ensure you have the latest CUDA drivers; on Apple Silicon, Metal Performance Shaders (MPS) are used automatically. If you’re on CPU-only, consider using a smaller model (like Stable Diffusion 1.5 instead of XL) or reducing image resolution. Docker Model Runner also allows you to adjust the inference backend settings—you can experiment with different schedulers or precision modes. For most users, the default settings work well. Keep an eye on memory usage; if you run out of RAM, the system may slow down or crash. Monitoring with docker stats helps.

9. Troubleshoot Common Issues

Even with a smooth setup, you might encounter hiccups. The most common issue is insufficient memory—if the model fails to load, try closing other applications or using a smaller model. Another issue is firewall or port conflicts; ensure port 8080 is free. If Open WebUI doesn’t launch, check that Docker Model Runner is correctly installed by running docker model version. For GPU errors, verify your drivers support CUDA or MPS. The Docker Model Runner documentation and community forums are great resources. Remember, since everything is local, you have complete control to tweak and debug.

10. Expand Your Setup: More Models and Features

Once you’re comfortable, you can explore beyond Stable Diffusion. Docker Hub hosts other DDUF models, including different versions (SDXL, SD2, etc.) or fine-tuned models for specific styles. You can also run multiple models side by side by launching separate instances on different ports. Open WebUI supports multiple backends, so you could even connect to a remote OpenAI API alongside your local model. For advanced users, consider integrating with image editing tools or automation scripts via the API. The possibilities are endless—and they all stay private on your machine.

Congratulations! You’ve built your own private AI image generator. No cloud subscriptions, no data leaving your computer, and no arbitrary content filters. Whether you’re a designer, hobbyist, or just curious, this setup gives you the freedom to create without limits. Start prompting and enjoy your new local art studio.