Who Text To Speech is for#
Portuguese-language readers reviewing long documents
Select text from any email, contract, PDF, or article and listen while multitasking. The floating player stays out of the way and closes automatically; the shortcut works without switching apps.
Skip if:
Text is primarily in English or another non-Portuguese language. Rafael and Piper both speak with a Brazilian Portuguese accent, which sounds unnatural on non-Portuguese text.
Privacy-conscious Mac users on restricted networks
Piper runs entirely on your Mac: after the initial model download, no text, no audio requests, and no analytics leave the device. Works on air-gapped machines and meets strict data handling requirements without any server configuration.
Skip if:
You need voices in multiple languages or want a larger voice library. Piper offers one Brazilian Portuguese voice; if voice variety is the priority, a managed service covers more ground.
Developers who want to run their own TTS server
The server API is a single POST endpoint with a documented request and response shape. Point the app at your own server URL via config.json or the TEXT_TO_SPEECH_SERVER environment variable to swap the backend entirely.
Skip if:
You have no interest in server customization. The public server works without any configuration; self-hosting is optional, not required.
The problem it solves#
Reading long documents, emails, and PDFs on screen strains the eyes, and listening while doing something else is often more efficient. Mac's built-in text-to-speech voices sound robotic, and switching between apps to use a dedicated reader breaks the reading flow. Most natural-voice alternatives are subscription-based commercial apps that require an account, charge monthly for quality voices, and send all text to a cloud service regardless of how sensitive the content is.
For Brazilian Portuguese speakers the pain is sharper: most text-to-speech tools treat Portuguese as an afterthought, offering at most one generic voice with poor intonation. Getting a natural-sounding voice for everyday reading on macOS has meant either accepting a subscription or accepting a robotic voice.
How it solves it#
Offline playback with Piper
The Piper voice model runs entirely on your Mac after a one-time download of about 63 MB from Hugging Face. No text leaves the device; playback works without any network connection. Rafael, the online voice, fetches audio from Fish Audio over HTTPS and does not store the text on the server.
System-wide keyboard shortcut
The default shortcut Cmd+Shift+S reads whatever text is selected in any Mac app: email clients, PDF viewers, browsers, text editors. The app copies the selection, restores your clipboard to its previous contents, and starts reading without interrupting the foreground app. The shortcut is configurable in the settings window.
Floating player with navigation controls
A small floating player appears on screen while reading, showing the word being spoken, and closes automatically when the selection ends. Esc stops playback; Space pauses and resumes; left and right arrows step between paragraph-sized parts; minus and plus keys adjust speed without opening the settings.
Adjustable reading speed from 0.5x to 2x
Speed can be set to half, normal, 1.5x, or double rate. On long documents, faster playback shortens a listening session without losing the thread; slower works for dense technical text or for following along with a printed page.
Paragraph-aware text splitting
Long selections are split into paragraph-sized parts, three processed at a time, so the first part starts playing within seconds even on a several-thousand-word document. Parts are read in order, and navigation between them is available with the arrow keys in the floating player.
Strengths and trade-offs#
Strengths
- Fully offline voice via PiperUnlike Speechify or NaturalReader, which send all audio requests to their cloud servers, Text to Speech includes Piper: a voice that runs on your Mac with no network connection after the initial model download. This matters for reading sensitive documents, working on restricted networks, or when connectivity is unreliable.
- MIT license with no subscription feesThe app is MIT licensed and free to download and run. It works without an account, a trial, or a paid tier. Speechify's natural voices cost $139/year or more; Rafael and Piper are both free and included in the open source release.
- No account, no telemetry, no Dock iconThe app requests one macOS permission (Accessibility, for reading the selection) and collects no analytics. It runs silently in the menu bar without a Dock icon, and settings are stored locally in ~/Library/Application Support/.
- Signed and notarized by AppleThe release build is signed with Developer ID and notarized by Apple, so macOS opens it without a security warning on any supported Mac. The source code is MIT licensed and available on GitHub for independent verification.
Trade-offs
- -Brazilian Portuguese voices onlyBoth Rafael and Piper speak Brazilian Portuguese. Reading text in English, Spanish, or other languages is possible, but the accent is Portuguese throughout. Users who need natural-sounding voices in other languages should evaluate Speechify or NaturalReader, which offer larger multilingual voice libraries.
- -Apple Silicon (M1 or newer) requiredThe published binary targets macOS 11 or newer on Apple Silicon only. Older Intel Macs are not supported by the pre-built release; users on Intel hardware would need to build from source using Tauri and Rust.
- -Online voice relies on third-party uptimeRafael sends text to the Fish Audio API. When Fish Audio or the intermediary server is busy, a part may fail; the app retries automatically, but extended outages will interrupt playback. The Piper offline voice is not subject to network availability and continues working when the public server is unreachable.
Text To Speech vs alternatives#
Text to Speech vs Speechify
Speechify is one of the most widely used paid text-to-speech services, available on iOS, Android, macOS, and as a browser extension. Its core value is a large library of AI voices across dozens of languages; Speechify Premium starts at $139/year for the best voices.
| Feature | Text to Speech | Speechify |
|---|---|---|
| License | MIT | Proprietary |
| Price | Free | $139+/year |
| Self-hosting | Optional (custom server) | No |
| Offline voice | Yes (Piper) | No |
| Languages | Brazilian Portuguese only | 30+ languages |
| Platform | macOS (Apple Silicon) | iOS, Android, macOS, Chrome |
Text to Speech is the better choice for Mac users reading primarily in Brazilian Portuguese who want offline capability and zero cost. Speechify is the better choice when you need multilingual support, cross-platform sync, or AI-driven features like reading speed training on mobile.
Text to Speech vs NaturalReader
NaturalReader provides a web app, a desktop app, and a Chrome extension with support for 20+ languages. Free accounts use basic voices; natural AI voices require a subscription starting around $9.99/month.
| Feature | Text to Speech | NaturalReader |
|---|---|---|
| License | MIT | Proprietary |
| Price | Free | Free basic / $9.99+/month |
| Offline playback | Yes (Piper) | No |
| Languages | Brazilian Portuguese only | 20+ languages |
| Platform | macOS (Apple Silicon) | Web, Windows, macOS |
NaturalReader is the stronger pick for multilingual needs; it supports English, Spanish, French, and many others with natural AI voices across platforms. Text to Speech is the better fit for Brazilian Portuguese Mac users who want MIT-licensed open source code, an offline option, and no ongoing cost.
Quick start#
Run the setup script to install on macOS 11 or newer on Apple Silicon.
```bash
curl -fsSL https://blob.ciromaciel.click/tools/text-to-speech/install.sh | bash
```What it's built on#
- Languages
- JavaScriptRust
FAQ#
Does Text to Speech work with any Mac app?
Yes. The app uses Accessibility permission to copy whatever text is selected in the foreground app, so it works in email clients, web browsers, PDF readers, word processors, and terminals. If an app does not support standard text selection, the shortcut will not capture text from it.
Does the Piper voice work without an internet connection?
Yes, once the voice model is downloaded. Piper downloads about 63 MB from Hugging Face the first time you select it, then runs entirely on your Mac. After that, no text and no audio requests leave the device.
Is Text to Speech free? Does it require a subscription?
The app is MIT licensed and free to download and use. There is no account, no trial period, and no paid tier. Both Rafael (online via Fish Audio) and Piper (fully offline) are included at no cost.
Can Text to Speech read documents in English or Spanish?
It can read text in any language, but both available voices speak Brazilian Portuguese, so non-Portuguese text is read with a Portuguese accent. For native-sounding English or Spanish voices, tools like Speechify or NaturalReader offer a wider multilingual library.
Does my text get stored by Text to Speech or Fish Audio?
When using Rafael (the online voice), your text travels over HTTPS to the tool's server, which forwards it to Fish Audio to generate audio and returns the result to your Mac. Neither service stores the text. When using Piper (the offline voice), text never leaves your Mac at all. No account is created and no analytics are collected by the app.
Similar open-source tools#
OpenClaw
392,257 GitHub starsSelf-hosted AI assistant for WhatsApp, Telegram, and more
Alternative to
Muse
Inbox Zero
12,602 GitHub starsAI email assistant that drafts replies and clears your inbox
Alternative to
Fyxer AI
rakazo
4,431 GitHub starsAI teammates you own: your keys, your model, your machine.
Alternative to
Grok Bot
paperclip
103,487 GitHub starsSelf-hosted AI agent management with org charts and budgets
Alternative toGrok Bot
Speech To Text
2 GitHub starsLocal speech-to-text that types where your cursor is
Alternative to
Wispr Flow
Voice Assistant
2 GitHub starsVoice interface for Claude Code with local Whisper and spoken answers
Alternative to
Superwhisper