FralRater

Readability that tells you who your text is too hard for — and shows you the fix.

FralRater is a fully offline desktop app. It scores your writing for specific reader personas, maps exactly where readers get lost, and rewrites the hard parts in front of you. Your text never leaves your machine — no network, no accounts, no AI.

Hemingway tells you where your writing is hard. FralRater tells you who it's too hard for — and shows you the fix.

Screenshots

---

Why it's different

Features

• Analyze — Flesch-Kincaid grade, reading ease, Gunning fog, SMOG, jargon density, long-sentence and passive-voice ratios, hard-word percentage.
• Audience — 0–100 fit score for 8 reader personas + per-paragraph confidence bars.
• Lost Reader map — SVG difficulty curve, one point per sentence.
• Rhythm heatmap — sentence-length variance at a glance.
• Rewrite — passive→active flip, buried-lead fix, jargon swaps, sentence splitting. Deterministic, rule-based, offline. Sentence-by-sentence diff with per-100-word jargon density before/after.
• Compare — two-version radar chart + head-to-head metric table.
• Voice note — dictate into the Analyze box (Chrome/Chromium Web Speech).
• Right-click menus — Cut/Copy/Paste in the desktop app.

Download / Install

Prebuilt releases are attached to each GitHub Release:

📖 Full user manual: see MANUAL.md.

Run from source

npm install     # one-time
npm start       # opens the FralRater window

Or:

./start.sh      # Linux/macOS
start.bat       # Windows (double-click works)

Requires Node 18+.

Build the release artifacts

npm install

# Linux artifacts: AppImage + .deb + .rpm
npm run package:linux

# Windows installer: FralRater Setup x.y.z.exe
npm run package:win

# Windows portable: FralRater-portable-x.y.z-win-x64.zip (unzip-and-run .exe)
npm run package:portable

# All of the above
npm run package:all

Easiest: push a tag — GitHub Actions builds everything.

git tag v0.1.0 && git push origin v0.1.0

The workflow in .github/workflows/release.yml runs on GitHub runners (Linux + Windows), builds all five artifacts, and publishes them as a GitHub Release automatically. Watch it under the repo's Actions tab.

Manual local builds (fallback)

npm install
npm run package:all   # outputs land in dist/

• Linux (AppImage, .deb, .rpm): build on Linux. .rpm additionally needs sudo apt-get install rpm (for rpmbuild).
• Windows installer (.exe): build on Windows, or on Linux with Wine (sudo apt-get install wine64).
• Windows portable (.zip of the .exe): cross-builds from Linux without Wine.

Architecture

Zero runtime dependencies. The "app" is a static bundle in public/ served by a 60-line Node server (server.js) on a free local port. Electron (electron/main.js) owns the window, spawns the server, and kills it on close. All analysis lives in public/lib/ as pure functions:

The lib/ layer is a clean seam: swap in a different engine (or LLM) without touching the UI.

Security model

• contextIsolation: true, nodeIntegration: false
• Content-Security-Policy: default-src 'self'
• Navigation locked to the local server; external links open in the OS browser
• Single-instance lock

Privacy

No analytics, no telemetry, no network calls. Voice notes use the underlying browser engine's Web Speech API.

License

MIT

Code of Conduct

This project follows the Contributor Covenant Code of Conduct.

---

🙏 Support open-source tool development

Your donation keeps this project maintained and funds new open-source projects, while supporting my CyberSecurity studies. Even a small amount makes a real difference. Thank you for supporting independent open-source work!

Made with care by Fralsare.