Making pictures

GlirOS can make a picture from a description: write "draw a fox in the snow" or "lag et bilde av en fjord", or press the picture button by the message box first and just describe it. Say "wide" or "landscape format" ("liggende") or "portrait format" ("stående") for another shape. Right after a picture, "make it blue" or "same but at night" changes it, and the regenerate button makes a new one from the same description. Hover the picture to save it. In the browser, pictures are made with your own OpenAI or Google Gemini key (Settings > Cloud models): only the description goes to that company, GlirOS asks once per company first, and the company charges your account. The desktop app can also make pictures on the PC itself with a picture model downloaded in Settings > Pictures (Z-Image Turbo): nothing leaves the PC; a graphics card with 8 GB or more makes one in seconds, the processor takes minutes. Settings > Pictures lists every picture model with an on/off switch (the dropdown shows the ones switched on) and the usual shape. Both versions also have Janus-Pro 1B, which makes 384×384 pictures on this device's graphics card (WebGPU; a 2.4 GB download the first time, asked first). Google's free Gemini tier has no picture quota: that needs billing on the key. The picture tool in the side panel ("open the picture tool" / "åpne bildeverktøyet", or Edit on a picture) has its own description box and grows with the controls you ask for ("add brightness and crop"): model, shape, style, what to avoid, seed, and edits like brightness, contrast, colour, black and white, sepia, blur, rotate, flip, crop and zoom; the result goes into the chat, to a download or into the project folder. The picture tool also opens pictures you added (Edit on the picture) and picture files in the project folder (click them in the Files panel); "Centre zoom" crops into the middle, "Zoom (view)" only enlarges the view. When a model fails, the next one takes over by itself (models on the PC, then cloud, then Janus-Pro last). In a project with a folder, "index and describe the pictures in the folder" ("lag en oversikt over bildene i mappa") has GlirOS list every picture and a model that can see pictures describe each one; pictures.xml and gallery.html are saved in the folder and shown in the side panel (with a cloud model GlirOS asks first, since the pictures go to that company). Chat models don't get the pictures GlirOS made, only the answer's words.

This guide is the same text GlirOS itself uses to answer questions about GlirOS.