Getting Started

New to ScoreSight? Follow these two pages in order:

  1. Installation: download and install ScoreSight on Windows, macOS or Linux, and grant the camera and screen-recording permissions it needs.
  2. Quickstart: open a video source, draw boxes over the score and clock, check the readings and send them to your graphics.

System requirements

  Requirement
Windows Windows 10 (version 1809 or newer) or Windows 11, 64-bit (x64). The installer checks this
macOS macOS 15 (Sequoia) or newer. Separate downloads for Apple Silicon (M1 and later) and Intel Macs
Linux 64-bit x86_64 distribution with glibc 2.35 or newer (for example Ubuntu 22.04 or newer, Debian 12 or newer) and an X11 or Wayland desktop
CPU / RAM Any recent dual-core CPU and 4 GB RAM. CPU use grows with the number of boxes and the Detections / s setting (see Performance Tuning)
Network Optional. Needed only for NDI, URL/RTSP sources, network outputs and the update check

ScoreSight 0.1.x and older ran on older macOS versions. Starting with 0.2.0, the bundled OCR engine and Qt framework require macOS 15 or newer.

Key concepts

Source
Where the picture comes from: a camera or capture card, an NDI feed, a URL (HTTP/RTSP) stream, a screen or window capture, or a video file. See Video Sources.
Box (field)
A rectangle you draw over one value on the scoreboard, such as “Home Score” or “Time”. Every box has a name and its own recognition settings. See Detection Boxes & Templates.
Reading (result)
The text ScoreSight recognized in a box during one detection cycle, together with its state: new, unchanged, failed filter or empty. New and unchanged readings are the good readings that outputs use.
Output
A destination for the readings: text files, the built-in HTTP server, OBS, vMix, overlays.uno or your own HTTP endpoint. See Outputs.
Profile
A named, independent set of settings, so you can run several ScoreSight windows side by side. See Multiple Instances & Profiles.

Table of contents