pim

A local Electron app that reads Markdown documents aloud using Kokoro — a high-quality, open-weight neural TTS model that runs entirely on your machine.

Features

• Open any folder of .md files and play them as a playlist
• Neural TTS via Kokoro-82M (runs locally, no API key needed)
• Sentence-by-sentence playback with live highlight in the rendered document
• 28 English voices (American & British)
• Speed control (0.75× – 2×)
• Play / Pause / Stop / Prev / Next controls + keyboard shortcuts
• Auto-advance to the next document when one finishes

Requirements

• Node.js 18+
• Windows, macOS, or Linux

Setup

npm install

The Kokoro model (~85 MB) downloads automatically from Hugging Face the first time you press Play and is cached locally for all future runs.

Run

npm start

Windows users: run from PowerShell or Command Prompt, not Git Bash.

Usage

1. Click Open Folder in the sidebar and select a directory containing .md files.
2. Click a file to load it, then press Play (or Space).
3. On first play the model downloads — a progress bar shows in the status strip.
4. Once loaded, audio streams sentence-by-sentence as it is generated.

Keyboard shortcuts

Key	Action
Space	Play / Pause
Escape	Stop
→	Next document
←	Previous document

Project structure

pim/
├── main.js          # Electron main process — window, IPC, file system
├── tts-worker.js    # Node.js worker thread — Kokoro inference (off main thread)
├── preload.js       # Context bridge — exposes safe IPC API to renderer
└── renderer/
    ├── index.html
    ├── renderer.js  # UI logic, audio queue, Web Audio API playback
    └── styles.css