NVDA add-on for Windows
VisionMate
AI image descriptions for the NVDA screen reader, running locally on your computer.
What it does
VisionMate describes images with AI and speaks the result through NVDA. Describe the current element, the whole screen or a clipboard image. Press once for a quick spoken description, twice for a detailed one in a window you can copy from. Automatic descriptions of unlabeled images on web pages and documents are optional and off by default.
Who it is for
Blind and low-vision people who use NVDA on Windows and want to know what is in an image: a photo, a chart, a meme, an unlabeled button's icon, or a screen that has no accessible text at all.
Features
- Describe the current element, the whole screen, the current window or a clipboard image
- Quick spoken description on a single press; detailed description in a copyable window on a double press
- Accessible Overlay (NVDA+Alt+O): turns a completely inaccessible window, app or game screen into an accessible overlay you can interact with
- Names for unnamed buttons from their icon (NVDA+Alt+B), or give them your own names (NVDA+Alt+N)
- Ask follow-up questions about an image
- Works with NVDA's screen curtain on
- Optional automatic descriptions of unlabeled images, off by default
Private and offline
After a one-time model download of about 1.5 GB, VisionMate works fully offline. No account, no API key, no cloud upload: your images never leave your computer.
Requirements
- Windows 10 or 11 (64-bit)
- NVDA 2024.1 or later
Get VisionMate
VisionMate is in the NVDA Add-on Store: in NVDA, open Tools, Add-on Store, Available add-ons, and search for "VisionMate". Updates arrive through the Add-on Store. The source code is publicly available on GitHub under GPL v2 or later.