AI chatbot for conversation, work, research, coding, and content creation.
Z-Image-Turbo
FreeStarting price
0Popularity
About Z-Image-Turbo
Z-Image-Turbo is an AI image generator that operates entirely within the user’s web browser using WebGPU acceleration. The model is downloaded once from Hugging Face and cached locally, so no data is uploaded to servers and prompts remain private. Generation occurs on the user’s device, with the browser handling tokenization, text encoding, denoising, and image decoding. The process requires a compatible GPU with WebGPU f16 shader support, as the model weights total approximately 5.5 GB and exceed memory limits of WebAssembly backends. Users can control resolution, denoising steps, and random seeds, enabling reproducible outputs without watermarks or quotas. The tool supports manual model loading and optional model mirroring for restricted networks. Once loaded, generation proceeds locally without further downloads.
Key features
- Local AI image generation without cloud processing
- WebGPU acceleration for faster denoising
- Adjustable resolution, steps, and seed control
- Manual model loading and caching
- Optional model mirror for restricted networks
- Direct download of cached model files
- No watermarks or quotas on generated images
- Private prompt and image handling
Use cases
- Prototyping visual concepts from text descriptions
- Generating images for personal or non-commercial use
- Testing AI image generation locally without data exposure
Pros
- Runs entirely in-browser with no server-side processing
- No account, queue, or watermark required
- GPU-accelerated via WebGPU for faster generation
- Model cached locally after first download
- No prompts or images are uploaded to external servers
Cons
- Requires a compatible GPU with WebGPU f16 support
- Initial download of ~5.5 GB model files
- Age restriction (18+) enforced
- Limited to browser-exposed GPU selection on Windows
Frequently asked questions about Z-Image-Turbo
Where does the model come from?
The page downloads the INT4 ONNX build of Z-Image-Turbo directly from Hugging Face, which serves it with permissive cross-origin headers. The 73hi server never stores or forwards the model.
Why is the download so large?
The text encoder and the transformer together are about 5.5 GB even in 4-bit quantization. They are cached in the browser after the first run.
Do I need a GPU?
Yes. The transformer weights alone are 3.44 GB, and the WebAssembly backend is built for wasm32, whose entire memory is capped at 4 GB, so the model does not fit there. WebGPU streams the weights to the graphics card.
How do I select a specific GPU for the browser?
On computers with multiple GPUs, you may need to select a compatible GPU for the browser in system graphics settings and fully restart the browser. The browser must expose that GPU with WebGPU f16 shader support.
Can I use Z-Image-Turbo without an internet connection after the first download?
Yes, once the model is downloaded and cached locally, generation proceeds entirely offline. The browser handles tokenization, text encoding, denoising, and image decoding without further downloads.
What happens to my prompts and generated images?
The prompt and every picture generated stay in the browser. The 73hi server never receives, reviews, or stores them, ensuring privacy and no content filtering.
