AI image generator
The AI runs on your graphics card in this tab. Your words and pictures are never uploaded.
Your pictures appear here.
Janus-Pro-1B · DeepSeek Model License……
…
Results
How to use this tool
Make a picture from a text description in this tab.
Describe it
Write what should be in the picture, or tap an example. Subject, place, light and style words help.
Pick style and size
Choose a style, how many pictures, and 768 or 1536 px if you want the AI to enlarge the result.
Generate and download
The first run downloads the model once. Click a result to view it large, then save it as PNG.
FAQ
How does the AI image generator work?
Janus-Pro-1B, an open model from DeepSeek, reads your text and draws a 384 × 384 picture token by token. Real-ESRGAN can then enlarge it to 768 or 1536 px. Both run in this tab. Prompts in other languages are first translated to English by OPUS-MT (Helsinki-NLP, CC BY 4.0 / Apache-2.0), about 110 MB per language, also in this tab.
Is it free? Is there a watermark?
Free, no account, no daily limit and no watermark: the work runs on your own device, so there is no server cost to charge for.
Can I use the images commercially?
Yes. The DeepSeek Model License allows commercial use. It forbids some uses, such as illegal content, content that harms minors, disinformation and harassment, so those are not allowed here either.
Why is the first image slow?
The model downloads once (about 1.5 GB) and is kept in your browser. After that, an image takes from a few seconds to about a minute, depending on your graphics card.
Why does it say WebGPU is needed?
Drawing needs a graphics card through WebGPU (recent Chrome, Edge or Safari). On the CPU alone one image would take many minutes, so the page does not start it.