Speak Helper

Read selected text aloud with one click — a lightweight desktop text-to-speech (TTS) assistant for Windows, macOS, and Linux.

English · 中文

Also known as: clipboard TTS, read-aloud tool, screen text speaker, OCR-to-speech, OpenAI TTS client, Edge TTS desktop app, 文字转语音, 剪贴板朗读, 选中朗读.

Use cases

Speak Helper turns on-screen text into speech — useful whenever reading is tiring, difficult, or inconvenient.

Prefer listening over reading

Scenario	Example
Too tired to read	Listen to articles, emails, or chat messages after a long day
Long documents	Hear reports or PDF excerpts without scrolling line by line
Small / dense UI	Copy text from apps with cramped fonts and let it read aloud

Vision & reading support

Scenario	Example
Low vision	Hear selected text from browsers, PDFs, or desktop apps (complements—not replaces—full screen readers)
Eye strain / fatigue	Reduce screen time by listening to copied paragraphs
Elderly users	Family can help set hotkeys; copy a message or news snippet to hear it clearly
Dyslexia / reading difficulty	Audio reinforces written content from any app

Note: Speak Helper is a read-aloud assistant for text you copy or select. It is not a full screen reader (NVDA, VoiceOver). It works well alongside those tools or when you only need occasional listening.

Daily productivity

Scenario	Example
Multitasking	Listen while coding, cooking, exercising, or commuting
Language learning	Natural Chinese / English neural voices for listening practice
Screenshot & image text	Copy a screenshot → OCR extracts text → automatic read-aloud
Quick workflow	Ctrl+Alt+R on selected text — no browser extension required

Features

• Floating dock on screen edge — states: idle, speaking, paused, error
• Clipboard trigger — copy text → prompt bubble (ask mode) or auto speak
• Global hotkey — Ctrl+Alt+R reads currently selected text
• Ask / Auto modes — switch from tray menu
• Two TTS backends
  • Edge TTS (Microsoft neural voices, no API key)
  • OpenAI-compatible API (/audio/speech) — OpenAI, DashScope, SiliconFlow, self-hosted
• Clipboard image OCR — vision API extracts text, then TTS (PaddleOCR-VL, etc.)
• Streaming playback — sentence chunking, fetch-and-play pipeline
• Local audio cache — same text skips API calls
• Secure API key — system keyring + config file
• Single instance — Windows named mutex

Quick start

Install

cd speak_helper
uv venv
uv sync

Or with pip:

pip install -r requirements.txt

Run

uv run python -m speak_helper
# or: speak-helper  (after pip install -e .)

First run: right-click system tray → Settings → choose TTS backend and API key (if using OpenAI-compatible mode).

Default backends

Backend	API key	Notes
edge (default)	Not required	Uses edge-tts / Microsoft neural voices
openai	Required	Any OpenAI-compatible /audio/speech endpoint

See docs/configuration.md for all options.

Optional .env: copy .env.example if you want team defaults. Model fields in .env are optional — configure in ⚙ Settings. Priority: config.json → .env → defaults.

Keyboard shortcuts

Action	Shortcut
Speak selection	Ctrl+Alt+R
Confirm bubble	Enter
Cancel bubble	Esc
Pause / resume listening	Single-click floating dock
Replay last text	Double-click floating dock

Build executable

Windows

.\scripts\build_win.ps1
# Output: dist\SpeakHelper\SpeakHelper.exe

macOS

bash scripts/build_mac.sh
# Output: dist/SpeakHelper.app

Documentation

Topic	English	中文
Index	docs/README.md	docs/zh/README.md
Installation	docs/installation.md	docs/zh/installation.md
Configuration	docs/configuration.md	docs/zh/configuration.md
FAQ	docs/faq.md	docs/zh/faq.md
Use cases	docs/use-cases.md	docs/zh/use-cases.md
Architecture	docs/architecture.md	docs/zh/architecture.md

Tests

uv run pytest tests/

Project layout

speak_helper/
├── speak_helper/          # Python package
│   ├── main.py            # App controller & wiring
│   ├── config.py          # JSON config + keyring
│   ├── speech_service.py  # TTS (edge + OpenAI-compatible)
│   ├── ocr_service.py     # Vision OCR for clipboard images
│   ├── clipboard_watcher.py
│   ├── hotkey_service.py
│   ├── audio_player.py
│   └── ui/                # PySide6 widgets
├── docs/                  # User & contributor docs
├── scripts/               # Build scripts
├── tests/
├── pyproject.toml
└── config.example.json

Privacy

• Text is sent only to your configured TTS / OCR API (or Microsoft Edge TTS for edge backend).
• No third-party analytics or relay servers.
• API keys stored in OS credential store when possible.
• Local cache lives in the user config directory; clear from Settings.

Contributing

See CONTRIBUTING.md (中文).

License

MIT License — Copyright (c) 2026 Speak Helper contributors.

Search keywords

text-to-speech, TTS, read aloud, clipboard reader, global hotkey TTS, PySide6, Qt desktop, Edge TTS, OpenAI speech API, CosyVoice, MOSS-TTS, OCR to speech, PaddleOCR, vision API, accessibility, dyslexia, 文字转语音, 朗读助手, 剪贴板朗读, 全局快捷键, 识图朗读.