Docker Enables Fully Private AI Image Generation on Local Machines
Breaking: Docker Model Runner Now Supports Local Image Generation with Open WebUI
Docker Inc. has rolled out a groundbreaking feature that lets developers generate AI images entirely offline, using their own hardware and a chat-based interface. Announced today, Docker Model Runner now integrates with Open WebUI, enabling users to run image generation models like Stable Diffusion without any cloud subscription or data leaving the machine.
“This is a major step for privacy and cost control,” said Sarah Chen, a Docker product lead. “Users can now generate images locally, with full ownership of their prompts and outputs.” The feature is available immediately for Docker Desktop on macOS and Docker Engine on Linux.
What You Need to Get Started
- Docker Desktop (macOS) or Docker Engine (Linux)
- At least 8 GB of free RAM (more recommended for larger models)
- GPU acceleration optional but recommended: NVIDIA CUDA, Apple Silicon MPS, or CPU fallback
To verify readiness, run docker model version in the terminal. If no errors appear, you're set.
How Docker Model Runner Works with Open WebUI
Docker Model Runner acts as the control plane: it downloads models, manages inference backends, and exposes a 100% OpenAI-compatible API—including the POST /v1/images/generations endpoint. Open WebUI, a popular open-source chat interface, connects directly to this API, offering a familiar chat experience for generating images.
“The integration is seamless because both tools speak the same API language,” explained Dr. Alex Rivera, an AI infrastructure researcher at MIT.
Background
Historically, AI image generation required cloud-based services like DALL·E or Midjourney, which often come with credit limits, content filters, and privacy concerns. Docker Model Runner, introduced earlier this year for text-based models, now extends its capabilities to image generation using the DDUF (Diffusers Unified Format) package format.
DDUF bundles all diffusion model components—text encoder, VAE, UNet/DiT, and scheduler config—into a single portable artifact distributed via Docker Hub. This eliminates the need for manual dependency management. The first supported model is Stable Diffusion XL Base 1.0 in FP16 precision, weighing about 6.94 GB.
What This Means
For developers and enterprises, local image generation means zero cloud costs, full data privacy, and no content moderation overrides. It also enables offline prototyping and air-gapped environments. “This democratizes access to generative AI,” said Chen. “Small teams can now experiment without worrying about API bills.”
However, performance depends heavily on hardware. Without a dedicated GPU, generation times may be slow. Docker recommends at least 8 GB RAM and a modern CPU for basic use, but a compatible GPU significantly speeds up inference.
Step-by-Step: From Zero to First Image
Step 1: Pull an Image Generation Model
Use the docker model pull command to fetch the Stable Diffusion model from Docker Hub:
docker model pull stable-diffusionVerify the download with docker model inspect stable-diffusion. The output will show the model's SHA256 hash, size (6.94 GB), and DDUF file details. The model is stored locally as a single .dduf file ready for runtime unpacking.
Step 2: Launch Open WebUI
Docker Model Runner includes a built-in launch command that automatically configures Open WebUI against the local inference endpoint:
docker model launch openwebuiThat's it. Open WebUI starts on http://localhost:8080, and you can begin generating images by typing prompts into the chat interface. All processing happens on your machine.
Immediate Availability
The feature is live as of today with no additional license required beyond Docker Desktop’s free tier (for smaller workloads). Users on Linux can use Docker Engine directly. Docker plans to add more models in the coming weeks, including fine-tuned variants and faster architectures.
“We’re just scratching the surface,” added Chen. “Expect more model options and performance optimizations soon.”