Hey PH! 👋 I'm the maker of 73hi, a collection of free browser tools that run entirely on your device. Why I built this ---------------- Every "free" AI image generator I tried had the same catch: your prompt goes to someone else's server, gets logged, gets rate-limited, and comes back with a watermark. I wanted to flip that completely — so I packaged Z-Image-Turbo as INT4 ONNX and shipped the entire diffusion pipeline to run client-side with WebGPU. How it works ------------ - First visit downloads ~5.5 GB of model files straight from Hugging Face (not from my server) and caches them in your browser - Your prompt is encoded, denoised and decoded 100% on your own GPU — the 73hi server never sees a single character - Full control: 512/768/1024 resolution, steps, guidance and seed, so any image is exactly reproducible - PNG download, no watermark, no account, no daily limit Honest requirements ------------------- - Chrome or Edge 130+ with a GPU that has WebGPU f16 support - The first load takes a few minutes (it's 5.5 GB) - 1024×1024 needs a lot of VRAM — start with 512×512 if you're unsure, 768×768 is the sweet spot on most cards One thing I want to be transparent about ---------------------------------------- Because nothing is uploaded, I technically *can't* see or filter what the model produces. That's why the page requires an 18+ confirmation before generating, and why you're responsible for what you create. Privacy cuts both ways — I chose to give it to you. I'm happy to answer anything about WebGPU, ONNX quantization, or what it took to squeeze a diffusion model into a browser tab. Ask away!

73hi — AI Images in Your Browser

AI image generation that never leaves your browser

AI image generation that never leaves your browser