FALSEOCR
FalseOCR is an open-source, lightweight Optical Character Recognition (OCR) capture utility designed explicitly to empower live esports production, tournaments, and broadcast overlays. Built on Electron and Node.js, this tool captures live video feeds from specific windows, screens, or cameras, and uses the Tesseract.js engine to extract text and bar values in real-time. By mapping specific regions of a game feed, broadcasters can automatically convert on-screen visual data (like player scores, gold, or health bars) into dynamic, accessible data. The extracted information is instantly served to a local network via an Express JSON endpoint or saved directly as individual .txt files, making it seamlessly compatible with professional broadcast software like OBS Studio, vMix, and custom real-time HTML/CG overlays.
100% FREE
Key Features
Advanced Live Capture & Mapping
Targeted Source Selection
Capture specific application windows, entire screens, or direct camera feeds using native Electron desktop capturers.
Visual Region Mapping:
Intuitive click-and-drag interface to draw bounding boxes directly over the live video preview.
Dual-Mode Reading
Text Mode Utilizes the OCR engine to read strings, scores, and names.
Bar Mode Analyzes pixel color spread to calculate percentage fills (ideal for dynamic health, mana, or progress bars).
Powerful Real-Time OCR Engine
Tesseract.js Integration
Runs a dedicated local worker to process text regions with minimal latency.
Custom Text Filtering
Optimize reading accuracy by forcing the engine to look for specific data types:
All Characters: For player names and general text.
Numbers (KDA/Score): Cleans up misread characters (e.g., reading an "O" as a "0") and filters out non-numeric characters.
Currency (Gold/Money): Isolates numerical values while ignoring trailing text or icons.
Color Inversion Toggle
Adjust reading parameters for bright/white text versus dark text to maximize contrast and accuracy.
Custom Tessdata Support
Load custom trained language models (.traineddata) for specific game fonts or languages.
Broadcast-Ready Data Outputs
Live Local JSON Endpoint
Hosts a local Express server (default Port: 14337) that updates and serves a live [ProfileName].json file, allowing remote web overlays to fetch real-time game states across the local network.
Direct .txt Output
Automatically saves and updates individual text files for every mapped label, perfect for direct integration into OBS text sources or vMix titles.
CORS Enabled
Allows external web applications and overlays from any origin or IP to fetch the live JSON data without security blocks.
Profile Management & Analytics
Import/Export Profiles
Save your exact region mappings for different games or tournaments and load them instantly via JSON profile files.
Live Analytics Dashboard
Monitor the health of your capture setup with real-time stats on Capture FPS, Engine Latency (ms), and the total number of areas being scanned.
Live Data Tree Preview
A built-in terminal view that shows exactly what data is currently being parsed and outputted by the engine.