Piper Document Reader

A simple desktop app that reads your documents aloud using Piper neural text-to-speech.

Load a Word document, Markdown file, or plain-text file, generate natural-sounding speech locally on your machine, and listen with full playback controls — play/pause, skip, seek, adjustable speed, and save to WAV. No cloud, no API keys, everything runs offline.

Markdown is cleaned automatically — headings, tables, and symbols become natural prose ready to speak.

---

Features

• Reads .docx, .md/.markdown, and .txt files — Markdown syntax (headings, badges, links, tables, code spans) is stripped automatically so it reads naturally.
• Offline neural TTS — powered by Piper voice models; no internet connection required.
• Full playback controls — play, pause, stop, rewind 10s, forward 10s, and a draggable seek bar.
• Adjustable speed — 0.5x to 2.0x presets.
• Save to WAV — export the generated audio to a file.
• Drag-and-drop friendly CLI — pass a file path as an argument to load it on launch.

Quick Start

Option 1 — Download the executable

Grab the latest PiperReader.exe from the Releases page. The voice model is bundled inside the .exe — no Python install, no separate download, fully offline out of the box. Just run it.

Option 2 — Run from source

git clone https://github.com/rainfantry/piper-reader.git
cd piper-reader
pip install -r requirements.txt
python piper_reader.py

Voice model

The released PiperReader.exe ships with the en_US-lessac-medium voice baked in — nothing to download.

When running from source, the app looks for the model in this order:

1. Bundled inside the executable (release builds only)
2. Next to the executable / piper_reader.py
3. Your home folder: ~/en_US-lessac-medium.onnx (+ .onnx.json)

Download voices from the Piper voices collection and place both the .onnx and .onnx.json files in one of those locations. To use a different voice, change MODEL_NAME near the top of piper_reader.py.

Usage

1. Launch the app (PiperReader.exe or python piper_reader.py).
2. Click Browse and pick a .docx, .md, or .txt file — or paste a path and click Load.
3. Review the extracted text in the preview pane.
4. Pick a Speed, then click Generate Audio.
5. Use the playback controls to listen, and Save WAV to export.

You can also launch with a file directly:

python piper_reader.py "C:\path\to\document.md"

Requirements

• Python 3.10 or newer
• piper-tts
• pygame (audio playback)
• python-docx (Word document support)
• tkinter (bundled with standard CPython on Windows)

Building the executable

The release binary is built with PyInstaller. To produce a self-contained build with the voice model baked in:

pip install pyinstaller
pyinstaller --onefile --windowed --name PiperReader \
  --collect-all piper --collect-all onnxruntime \
  --add-data "en_US-lessac-medium.onnx;." \
  --add-data "en_US-lessac-medium.onnx.json;." \
  piper_reader.py

• --collect-all piper is required — Piper bundles espeakbridge.pyd and the espeak-ng-data/ dictionaries; without them synthesis fails at runtime.
• The two --add-data flags embed the voice model so the .exe runs with no external files. Drop them for a smaller binary that loads the model from disk instead.

The result lands in dist/PiperReader.exe.

License

Released under the MIT License.