Open Generative AI

Generate AI images, videos, and lip-synced animations using over 200 models with a free, open-source, and filter-free desktop application.

Visit Site
Generate AI images, videos, and lip-synced animations using over 200 models with a free, open-source, and filter-free desktop application.

The gist

Open Generative AI is an open-source desktop application that acts as a unified studio for creative AI. It provides a graphical interface to generate images, videos, cinematic shots, and lip-synced animations using over 200 different models. Built by Anil-matcha, it serves as a free, self-hosted, and filter-free alternative to proprietary AI media generation platforms, giving users full creative control over their outputs without subscription fees or content guardrails.

What it does

  • Generate images from text or other images using over 50 models in the Image Studio.
  • Create videos from text prompts or by animating still images across more than 60 models.
  • Animate portrait photos or sync lips on existing videos from an audio source.
  • Build and run multi-step AI pipelines visually with a node-based editor.
  • Run select models locally on your own hardware, with support for Metal on Apple Silicon.
  • Control virtual camera settings like lens, focal length, and aperture for cinematic shots.

How it works

The tool is a desktop application for macOS, Windows, and Linux that connects to over 200 cloud-based AI models via the Muapi.ai API. Users input text prompts, reference images, or audio files and receive generated images or videos. It also supports local inference, allowing users to run certain models directly on their own hardware. The application is free and open-source, providing an alternative to subscription-based services, with your data remaining on your own machine.

Best for

This tool is ideal for creators and AI enthusiasts who want to experiment with a vast array of generative models for images and video through a single interface, without content filters. It's also a strong choice for users who prefer to run models on their own hardware for privacy and control.

Watch out for

The desktop application is not code-signed, so users on macOS and Windows must manually bypass security warnings during installation. Running the most powerful local models requires a separate, self-managed server with a compatible CUDA or ROCm GPU.