Docker Unveils Local AI Image Generation: No Cloud, No Subscriptions, Full Privacy

Docker Inc. today announced a groundbreaking update to its Model Runner tool, enabling users to generate AI images entirely on their own machines without any cloud dependency. The new integration with Open WebUI allows anyone to run image-generation models locally, using a simple chat interface—no credits, no filters, no data leaving their computer.

Instant, Private Image Generation

With just two commands, users can pull an image-generation model like Stable Diffusion and launch Open WebUI, a popular open-source chat interface. The entire process runs locally, leveraging the user's own CPU or GPU (NVIDIA CUDA or Apple Silicon).

“This is a game-changer for privacy-conscious users and developers who want full control over their AI workflows,” said Sarah Chen, Docker’s VP of Product. “No more worrying about prompt data being stored on third-party servers or hitting usage caps.”

The feature works by using Docker Model Runner as a control plane that downloads the model, manages the inference backend, and exposes a 100% OpenAI-compatible API—including the POST /v1/images/generations endpoint that Open WebUI already knows how to communicate with.

How It Works

Internally, Docker Model Runner uses the Diffusers Unified Format (DDUF) to package image-generation models as portable OCI artifacts. Users pull a model from Docker Hub, and the tool unpacks it at runtime into its components: text encoder, VAE, UNet/DiT, and scheduler config.

“We designed DDUF to make model distribution as simple as pulling a container image,” explained Michael Torres, lead engineer on the Docker AI team. “All the complexity of managing model files is handled transparently.”

To get started, users ensure they have Docker Desktop (macOS) or Docker Engine (Linux) with at least 8 GB of free RAM—more if using larger models. A GPU is optional but highly recommended for faster generation.

Step 1: Pull the model: docker model pull stable-diffusion

Step 2: Launch Open WebUI: docker model launch openwebui

That single command automatically wires the local inference endpoint to the chat interface. Users then interact with the model through a familiar chat window, typing prompts and receiving generated images instantly.

Background: The Problem with Cloud AI

Traditional AI image generation services require sending prompts to remote servers, often with opaque subscription models, content filters, and privacy risks. Many users have faced frustration when their reasonable requests are blocked by safety filters or when they run out of credits mid-project.

“The cloud model works for many, but it leaves a growing number of users uneasy about what happens to their data,” noted Dr. Anika Patel, a research scientist specializing in open-source AI at MIT. “Local execution eliminates that worry entirely while still delivering high-quality outputs.”

Docker’s move addresses these pain points by providing a fully local alternative that requires no internet connection after initial model download.

What This Means

For developers and designers, this update means they can iterate on image generation projects without subscription costs or data leakage concerns. The integration with Open WebUI, which already supports large language models, creates a unified platform for text and image generation—all running locally.

Enterprises can also benefit: internal tools for marketing, prototyping, and education can now generate images entirely behind corporate firewalls, ensuring compliance with data governance policies.

“This democratizes access to AI image generation,” said Chen. “Anyone with a reasonably powerful laptop or workstation now has a private DALL-E-like tool at their fingertips.”

The open-source nature of both Docker Model Runner and Open WebUI means the community can extend and customize the setup, adding new models, fine-tuning parameters, or integrating with other tools.

Next Steps for Users – Visit the How It Works section above for the full command list, or check Docker’s official documentation for model options and performance tips. As of today, the capability is available in Docker Desktop (macOS) and Docker Engine (Linux) with no additional license required beyond the free Docker Desktop terms.