Freegma — Free Local AI Studio for Developers

Many developers and creatives don't have access to powerful GPU hardware for AI generation. Cloud-based AI services are expensive, and free tiers are limited.

I built Freegma, a free local AI studio that runs on my own hardware and is accessible to anyone through the browser. It supports text-to-image generation (SDXL 1.0, with FLUX 1.1 coming soon), image-to-3D model conversion, and image-to-video generation.

The entire AI pipeline runs locally using ComfyUI with custom workflows I built for each generation type. The backend is a custom script that runs on my PC and automatically loads the required AI models onto the GPU when a user starts a generation.

To make it publicly accessible, I set up a Cloudflare Tunnel that connects my local machine to the Vercel-hosted frontend. When someone visits Freegma, their request travels through the tunnel to my hardware, where the actual AI processing happens.

Text to Image: Generate images using SDXL 1.0 with full control over steps, CFG scale, resolution, and seed

Image to 3D: Convert 2D images into 3D models

Image to Video: Animate static images into video clips

Zero cost for users: Everything runs on my hardware, completely free

The frontend is built with Next.js and deployed on Vercel. The backend runs as a local Python service with ComfyUI orchestrating the AI models. Cloudflare Tunnel handles the secure connection between the public frontend and my local GPU server.

This is a personal project built in 2 days, offered for free to help developers and creatives who don't have access to expensive GPU hardware.