AI Image GeneratorBeta
Describe a picture and an open AI model draws it on your own device — no account, no credits, and nothing you type is sent anywhere.
Describe a picture and an AI model draws it on your own device. The model downloads once (about 1.7 GB) and is kept in your browser; your prompts and the pictures never leave this tab.
You need a recent browser, some free disk space, and patience on the first download. A GPU (WebGPU) makes everything many times faster and is used automatically when this device passes a real GPU test; without one, everything runs on the CPU.
Choose a model and press Download & load to begin.
The first load downloads the weights into this browser's storage. Later visits reuse them and start in seconds.
Resource usage no model loaded
Browsers do not report CPU or GPU load to a web page, so nothing here is invented. CPU load is measured two ways and shows the higher: how much of each second the engine spends blocking its own worker, and how much slower a fixed piece of arithmetic runs right now than on an idle machine — which is what a multi-threaded engine looks like from here. Memory is the JavaScript and WebAssembly heap, which only Chromium-based browsers expose. GPU utilisation and video memory are not available in any browser.
Generation settings
Speech to text, text to speech, music and images share one model slot, so loading one replaces the previous one — your chat model stays where it is.
Transcribe audio
Speech to text, text to speech, music and images share one model slot, so loading one replaces the previous one — your chat model stays where it is.
Speech to text, text to speech, music and images share one model slot, so loading one replaces the previous one — your chat model stays where it is.
Make music
MusicGen is slow without a GPU — measured on a plain CPU, about 50 seconds of computing for 2 seconds of music. Start with a short clip.
Speech to text, text to speech, music and images share one model slot, so loading one replaces the previous one — your chat model stays where it is.
Load any model from Hugging Face
Search the Hub, or paste a repository id. Each weight variant is listed with its real download size and whether it is likely to fit on this device.
Downloaded on this device
Models are stored by your browser and count towards its storage quota. Delete any of them here at any time.
Saved conversations
Your chats are kept in this browser only, so you can pick one up on your next visit. They are never uploaded, and deleting models above does not touch them.
This device
Graphics drivers and browsers get updated. If GPU inference was switched off after a failure, you can ask for it to be tested again.
Log
🔒 Runs entirely in your browser — nothing you type is uploaded or stored on a server.
This image generator runs the AI model on your own computer or phone, inside this tab. Most free AI image sites send your prompt to their servers, count it against a daily allowance and keep a copy. Here there is no server doing the drawing at all: your browser downloads an open model once, and every picture after that is computed on your own hardware.
Also on Txtset: chat with an AI model on the same device · turn your writing into a voice-over · write a short piece of music to go with it.
The model is Janus-Pro 1B, released by DeepSeek. It works differently from the Stable Diffusion family: a one-billion-parameter language model writes the picture as a sequence of 576 image tokens, and a small decoder turns those into a 384×384 image. That design is why it can run on an ordinary CPU at all — the work is the same kind of work a chat model does, which browsers now handle well.
Be realistic about what a model this size gives you. It is good at simple scenes, objects, animals, landscapes and illustration styles; it follows the gist of a prompt rather than every detail, and it is weak at hands, text inside the picture and crowded compositions, and 384×384 is small next to the big hosted generators. What you get in return is privacy, no account, no watermark and no limit on how many you make.
Every picture you make in a visit stays in a strip under the main one, so trying a prompt again never loses the first attempt, and each can be saved as a PNG. The model shares the studio with Local AI Chat, speech to text, text to speech and music generation — all on the same page, all on your device.
How to use
- Press Download & load. The model is about 1.7 GB and is stored in your browser, so this happens once; later visits load it from disk in seconds.
- Describe the picture you want in the prompt box — the subject first, then the style (“watercolour”, “photo”, “flat illustration”).
- Press Generate and watch the progress bar; it estimates the time left as it goes.
- Click Download PNG to keep the picture, or generate again — earlier results stay in the strip below.
Examples
Make a simple header image without licensing a stock photo or signing up for a credits plan.
Try a dozen takes on a colour palette or a setting, and keep the ones that work.
Describe something you would rather not type into a company's server — here it never leaves the tab.
Show how an image model actually works on a school laptop, with the network panel open to prove nothing is uploaded.