Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Voice, Speech & Multimodal

Parrot Review: Free Offline Text-to-Speech That Reads Selected Text Aloud

·7 min readBy Hans Kuepper · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

Warning: Parrot has had no update since 28 April 2026 (last release 26.2.4, 25 February 2026), so it may be unmaintained — check that it works on your system before relying on it. Parrot is a free, MIT-licensed desktop app that reads text you select in any program aloud, using a voice model that runs on your own computer. This review covers version 26.2.4 and is based on the public website, README and GitHub releases, not on hands-on testing.

Key Takeaways

  • Version reviewed: 26.2.4, released 25 February 2026; the newest code change on GitHub is from 28 April 2026, so the stale-project warning applies.
  • License and price: MIT per the repository; free, with no account.
  • Platforms: Windows, macOS on Apple silicon and Linux, as desktop apps.
  • Standout features: system-wide shortcut, streaming playback, pause and resume, local history.
  • Maturity: a small project (162 GitHub stars) that is a fork of the speech-to-text app Handy.

📍 In One Sentence

Parrot is an open-source desktop app that reads whatever text you select aloud, using a compact voice model that works offline.

💬 In Plain Terms

Think of it as a read-aloud button for your whole computer: highlight text in a browser, PDF or editor, press a key, and a natural-sounding voice reads it to you without any cloud service.

Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

Parrot is a free, open-source desktop app that reads any text you select aloud with a small neural voice model running on your own computer. This review covers what its website, README and GitHub releases document, what they leave out, and how it compares with other text-to-speech tools in the directory.

Parrot Review: Free Offline Text-to-Speech That Reads Selected Text Aloud

Get Parrot

Parrot ships as a double-click desktop app on all three desktop systems, so no terminal is needed. The README sends you to the website for the download; the installer files are also attached to the latest GitHub release.

Platform
Get It
macOS (Apple silicon)Parrot_26.2.4_aarch64.dmg
WindowsParrot_26.2.4_x64-setup.exe or .msi (ARM64 builds also listed)
LinuxAppImage, .deb or .rpm for x86-64 and ARM64
Websitetryparrot.vercel.app
Privacy policyNo separate privacy-policy page was found; the README and site describe local processing
Source codegithub.com/rishiskhare/parrot

The release lists only an Apple silicon disk image, so Intel Macs are not covered. On first launch the app asks you to download the voice model, which needs an internet connection once.

Parrot at a Glance

Attribute
Parrot
PlatformWindows, macOS, Linux
PriceFree
LicenseMIT
RunsLocally, on the CPU
Languages9 options (US and UK English counted separately)
Voice cloningNot stated; 54 preset voices
Streaming outputYes, audio starts before synthesis finishes
Usable on CPU onlyYes
Local API serverNo; command-line flags only

Attributes follow the text-to-speech comparison used in the Local LLM Software Directory. A missing statement on the site means unknown, not unsupported.

What Parrot Is

Parrot is a read-aloud utility, not a voice studio. The website calls it probably the most simple text-to-speech app: its only function is to read aloud whatever you select. The author says the goal is not the best app but the most forkable one, a small base for accessibility tools and voice experiments.

It is a fork of Handy, an open-source speech-to-text app by CJ Pais, and keeps its Tauri architecture with a Rust backend. The voice comes from Kokoro-82M, an open model with an Apache-2.0 licence on its Hugging Face page, run through the author's tts-rs library.

This review draws on the website, the README and the GitHub releases and commits. It involved no hands-on testing, so voice quality, speed and stability are not rated here.

How to Get Started

Setup is an installer download, a one-time model download, then a shortcut. The steps follow the README.

  1. 1
    Install the app for your system
    Why it matters: Download it from the [website](https://tryparrot.vercel.app/) or the [releases page](https://github.com/rishiskhare/parrot/releases/latest) and run the installer.
  2. 2
    Download the voice model on first launch
    Why it matters: Parrot prompts for the roughly 115 MB model; once it is stored, the README says the app works completely offline.
  3. 3
    Select text and press the shortcut
    Why it matters: The default is Option+Space on macOS and Ctrl+Space on Windows and Linux; a small overlay shows while audio plays, and every shortcut can be changed.
  4. 4
    Pause or resume while it speaks
    Why it matters: Option+P (Alt+P on Windows and Linux) pauses and resumes playback; this shortcut is only active during playback.

Features and Voices

The feature set is deliberately small. Everything below comes from the README and the website.

  • Works in any app. It reads selected text from browsers, editors, PDFs and terminals.
  • Streaming playback. Audio starts before the whole text has been synthesized.
  • 54 voices in nine language options. US and UK English, Spanish, French, Hindi, Italian, Japanese, Brazilian Portuguese and Mandarin; the voice follows your language setting or can be picked in Settings.
  • Controls. Playback speed, worker threads, an optional first-response speed-up and a floating overlay with pause controls.
  • History. Past utterances can be browsed, replayed, copied and deleted, with an entry limit and an automatic delete period.
  • Scripting. Command-line flags such as parrot --toggle-transcription and Unix signals let window managers and hotkey tools trigger speech, which is the recommended route on Wayland.

Privacy, License and Cost

The README says your text is processed on your device and never sent anywhere. It lists no cloud, no accounts and no internet requirement after the first model download. The website repeats that the text stays on your computer.

The code is MIT-licensed and the app is free with no subscription or API key, per the README. No dedicated privacy-policy page was found, and the README does not say which server hosts the model download. This review did not inspect network traffic, so these statements are the developer's own.

The History feature stores what Parrot has read on your machine; you can set an entry limit and an automatic delete period, or delete entries by hand.

Trade-Offs: Benefits vs. Limitations

Offline neural voice

What it means in real use:
Listen to text without a cloud service.
Limitation / caveat:
Voice quality is not rated here; the project links only an audio sample.

Works in any app

What it means in real use:
One shortcut for browsers, PDFs and editors.
Limitation / caveat:
Wayland needs a desktop-level shortcut that calls the command line.

Small and CPU-only

What it means in real use:
Runs on an ordinary computer.
Limitation / caveat:
No RAM figure is published.

Installers for all desktops

What it means in real use:
Download and open, no terminal needed.
Limitation / caveat:
macOS build is Apple silicon only.

Open source and free

What it means in real use:
MIT code you can fork and change.
Limitation / caveat:
No update since 28 April 2026; the author says feature requests may not make it into the main project.

Parrot vs. Alternatives

Most other text-to-speech tools in the directory are engines for developers, not read-aloud apps. Parrot is the one you install and use with a shortcut.

Parrot

Platforms:
Windows, macOS, Linux
Price and license:
Free, MIT
Role:
Read-aloud desktop app
Key difference:
Select text, press a key, hear it; no coding

Izwi

Platforms:
Windows, macOS, Linux
Price and license:
Free, MIT
Role:
Local voice runtime
Key difference:
Speech output plus transcription, cloning and an API
Articles about Izwi (2)

Platform, price, and feature details for third-party apps change frequently. Verify current specifics on each project's own page before deciding.

Compare in directory

Who Should Use Parrot

  • People who like to listen. Articles, long emails and documents become audio with one shortcut.
  • Readers with accessibility needs. The project frames the app as accessibility tooling that should not sit behind a paywall.
  • Tinkerers. The code is meant to be forked and extended.

Who Should Not Use Parrot

  • Intel Mac owners. The release lists only an Apple silicon macOS build.
  • Anyone who needs a maintained product. The stale-project warning at the top applies.
  • People who need voice cloning or an API. Neither is documented; look at the engines in the table above.
  • Users who need documented compliance. There is no privacy-policy page.

Frequently Asked Questions

Does Parrot need a GPU?

No. The README says the model runs on any modern CPU; it publishes no RAM or speed figures.

Is Parrot the same as Handy?

No. Handy is a speech-to-text app; Parrot is a fork that reverses the direction and reads text aloud.

Can I change the shortcuts?

Yes, all shortcuts can be changed in Settings. On Wayland, bind your desktop shortcut to parrot --toggle-transcription.

Does it keep a record of what it read?

Yes, a History page stores past utterances locally; you can limit the number of entries or delete them automatically.

Can it clone a voice or run as a server?

Neither is documented. It offers 54 preset voices and command-line flags only.

Verdict

Parrot is a good fit if you want a free, no-account way to hear selected text read aloud on your own computer: it installs like a normal app and does one job.

The caveats are maturity and transparency: no update since 28 April 2026, 162 GitHub stars, no privacy-policy page and no voice-quality or memory figures. Try it with a few documents before depending on it, and treat the stale-project warning as binding.

Sources

← Back to Power Local LLM

MCP