UnScreen. Open reader

A clearer way to follow along

Every screen share.
Readable, live.

Turn a shared screen into a table, code or text your screen reader can navigate. Stay with the conversation, even as the screen changes.

Start reading a shared screen See it in action
  • No presenter setup
  • Your screen reader, your pace
  • No account needed

Works alongside your meeting

  • zoom
  • Google Meet
  • Microsoft Teams
Built for NVDA, VoiceOver & ChromeVox

The screen, made accessible

A picture becomes a place
you can move through.

Real rows. Real headers. Your place stays put.
Explore a generated sheet rebuilt by the real model.

A generated spreadsheet in a simulated meeting window, read by the real model after lossy compression at 720p. UnScreen rebuilt every cell in the right place. The first spoken line is its real announcement; the others show what the table's headers let a screen reader say as you move.
Open a test spreadsheet to share (new tab)

Screen sharing should
include everyone.

In a study of 65 desktop and 94 mobile screen reader users, Zoom, Meet and Teams all rendered shared content as an image (Leporini, Buzzi and Hersh, ACM TACCESS 2023). A blind participant could tell a whiteboard was being shared, but not what was written or erased on it (PMC9483261). Screen reader OCR is the fallback, and users describe it as ranging from very successful to useless (WebAIM). It reads flat lines, so a spreadsheet loses its rows and columns.

A simpler way to stay in the meeting

Share. Rebuild. Listen.

  1. Share the meeting.

    Open UnScreen next to your meeting and press Start. Chrome asks which tab or window to read; pick the meeting. Your screen reader stays on UnScreen. The presenter installs nothing.

  2. Make it readable.

    When the shared screen changes and settles, UnScreen sends one picture to its engine. Gemini reads it into rows, columns, lines and headings. A second, independent read checks it, and anything the two reads disagree on is marked unclear.

  3. Follow every change.

    Read the result like a web page, with table navigation and headers. Changes are announced politely, once ("Acme Corp, Q3 revenue: $1.1M to $1.2M"), and your reading position never jumps.

Transparency, at every step

Measured, not claimed

10 test frames (sheet): shares inside a meeting window, sent at 720p with lossy compression. Compared with Tesseract OCR, given extra help to find table cells. Run on 24 Oct 2026 with gemini-3.6-flash.

MeasureUnScreenOCR baseline
Cells and lines read exactly99.6%41.3%
Wrong but stated as fact0.4%49.8%
Found values in the right row and column100.0%79.6%

Full results, method and what failed

Your screen. Your choice.

Pictures of the shared screen go to UnScreen's engine and on to Google's Gemini API to be read. UnScreen does not store them, and there are no accounts or analytics. The details.

Be part of the conversation.

Your next screen share can be a little more accessible.

Open the reader
UnScreen.Every screen. A shared understanding.

UnScreen is a student project built for the Taskforge hackathon, October 2026. Shared frames are sent to Google's Gemini API to be read and are not stored by UnScreen. How your screen is handled.

Source publication is pending. This version is being validated locally.