You can create an AI image from Chrome while the image model runs on your own computer. OGAD (Off Grid AI Desktop) supplies the image engine, and its browser extension gives you a Create image mode beside the page where you are working.
Use it to draft a blog illustration or explore a visual idea without moving your prompt to a cloud image service. You need a supported local image model installed and selected in OGAD. The browser panel itself does not run the model.
Download OGAD | Desktop beta 122 release notes
What would you like to do with Off Grid AI?
Have a feature or use case you would like us to support? Tell us what you want to do and which device you use.
Write to support@getoffgridai.co, join our Slack community, or talk to us on Reddit.
Start with a small illustration brief
Suppose you are writing a post about working away from the office. You want a quiet desk scene that leaves space for a title. The useful first result is one image with the correct subject and composition, not a large batch of variations.
Write the visual brief before generating. Include the main subject, surroundings, light, style, and where empty space should remain. If a page is open beside the panel, repeat the facts that matter in the prompt rather than assuming that image generation will infer your complete design brief from the page.
Prepare the desktop image engine
Install the current OGAD beta and open the image model controls. Download a compatible package and select it. A working text model does not establish that an image model is ready. Image generation can also require more memory than a short text reply.
Try one ordinary text-to-image request on the desktop if you have not used that model before. This checks the engine before you add the browser connection. Use a modest image size and the model's normal settings for the first result.
Get the early-access extension, then pair it in Settings > Desktop. Compare the six words and approve the matching request in OGAD. Keep the desktop awake and connected while the image job runs.
Generate from Chrome
- Open the extension beside the page where you are working.
- Open the composer controls and select Create image.
- If the panel says to install an image model, return to the desktop's image settings.
- Enter the brief and send it.
- Wait for the actual image result. Inspect the picture before treating the job as complete.
Try this:
Create a wide editorial illustration of a small wooden desk beside an open window, a laptop, a notebook, and morning light. Use a calm green and warm grey palette. Keep the right third simple for a heading. No text, logos, or watermark.
This prompt describes visible features. It does not ask the model to understand an entire article or reproduce a specific brand's artwork. You can add the title later in your normal design tool, where its spelling and placement are under your control.
Review the result as an illustration
Check the composition first. Does the subject fit your intended crop? Is the requested empty area useful? Then inspect small details such as hands, cables, labels, and repeated objects.
If the image is wrong, change one part of the prompt. “Move the laptop to the left and keep the right side empty” is a clearer revision than adding ten style words. Keep the prompt with the result so you can tell which change produced which version.
An image in the conversation is the useful evidence of completion. A text reply that says “I created it” is not enough. If the job reports an error or the image is missing, check the job result on the desktop before retrying.
Keep generation and publishing separate
The extension creates a picture through OGAD. It does not automatically insert that picture into your blog editor, publish a post, or grant rights to third-party material shown in the result.
Save the image through the available image controls, then open the saved file. Check the dimensions and crop in the destination where it will be used. Add clear alternative text when you publish it.
This guide covers text-to-image generation. Editing a reference picture is a separate capability that depends on the chosen image model and request route. Do not assume that a model which generates images can handle every editing operation.
If the job does not finish
| Symptom | Check |
|---|---|
| Create image is unavailable | A supported image model is installed and selected |
| Text chat works but image generation fails | Desktop image engine readiness and available memory |
| The browser loses its connection | Desktop availability and the paired address |
| Only a text claim appears | Look for a real image result or a job error |
| The image ignores part of the brief | Shorten the prompt and change one requirement at a time |
With a local image model selected, the computation stays on the desktop. A remote image endpoint changes that condition. Downloads need internet initially; the local generation job does not need a cloud image service afterward.
Prepare OGAD and generate one illustration from Chrome. Save and inspect the file before you use it in an article.
Top comments (0)