Make pictures

Pictures turns a description into an image. You need a picture model first: open Get models → Pictures. Stable Diffusion 1.5 is small and quick; Stable Diffusion XL gives better results if your graphics card has 8 GB or more.

If you installed the Windows app or a chat-only install, the first visit to Pictures (or Videos, Read aloud, Transcribe or Music) shows One more step: picture, video, voice and music support. Press Install for a one-time download of the AI libraries, about 3 GB, matched to your graphics card. You can keep chatting meanwhile.

The Pictures page

Describe your picture

Type what you want to see and press Create. Be specific about the subject, the setting, the style and the light, for example:

an old typewriter on a wooden desk, a brass lamp glowing, rain on the window at dusk, impressionist oil painting

The model starts by itself the first time, then the picture appears above the prompt bar with a progress bar.

Options

Option	What it does
Model	Which picture model to use.
Shape	Square, Wide or Tall. Sizes are chosen to suit each model.
Quality	Quick, Balanced or Best. Better quality takes more steps, so it’s slower.
Pictures	How many to make at once.
Start from a picture	Restyle one of your own pictures. Prompt strength sets how far it can move from the original: low keeps it close, high changes more.

Under More settings:

Setting	What it does
Things to avoid	What you don’t want, for example blurry, extra fingers, text.
Steps	How many refinement steps to run. Leave empty to follow Quality.
Sampler	How the picture is drawn step by step. Automatic works well for most pictures.
Seed	The same seed and description give the same picture. Leave it on Random for variety.

After it’s made

Save downloads the picture.
Use again puts the same description and settings back in the prompt bar, so you can try a variation.
Delete removes it.

Everything you make is also kept in the Gallery, and in the outputs folder of your data folder.