Open Generative AI: local image generation setup

This tutorial explains how to set up Open Generative AI for image generation on your own hardware. The speaker describes it as an MIT-licensed, self-hosted studio with a bundled C++ inference engine and support for macOS, Windows, and Linux distributions such as Debian and Ubuntu.

The installation walkthrough covers Windows SmartScreen and macOS Gatekeeper warnings. On Mac, it shows the command xattr -cr /Applications/OpenGenerativeAI.app, followed by opening the app through Finder's right-click menu. These steps remove quarantine restrictions and override launch warnings.

Inside the app, the speaker skips the API key prompt and opens the local model settings, where the app downloads sd.cpp binaries for the detected operating system and CPU architecture. The stated memory guidance is 16 GB of RAM for Z image architectures, while 8 GB machines should use sd1.5 models with an engine memory requirement below 2 GB.

The image studio requires users to enable local execution beside the model selector. The speaker says leaving it disabled causes an API request and a 403 error without credentials. After setup, the tutorial describes offline rendering and shows how dragging an output onto the canvas starts an image-to-image workflow. Its claims of zero telemetry and no external tracking are the speaker's assertions, rather than independently verified privacy guarantees.