Changelog
A dated record of every published PiperRead version — what changed, and when.
What this page tracks
A dated record of every version of PiperRead that has actually been published, most recent first, written for a visitor rather than copied from the technical release notes. The project was first published on 11 January 2026; the GitHub repository has since been recreated for internal reasons, which does not change that date. The exact detail of each version — the full notes, the diff, the downloadable assets — lives on the GitHub releases page, which stays the canonical source; this page exists so a visitor who has no reason to open GitHub can still see, at a glance, what changed and when.
v0.4.1-beta — 13 October 2026
Packages for Fedora 44 and openSUSE Leap 16.0 (0.4.0-beta was built for Fedora 42 and openSUSE Leap 15.6). The Windows archive now includes the license texts of its components in a licenses folder, the exact version of each component and a written offer of their sources. The README and the manual page state what each platform has been through. No function changes.
v0.4.0-beta — 29 September 2026
Two ways to install, one voice window. On every distribution family (Debian and Ubuntu, Fedora, openSUSE Leap, Arch) you now choose: the piperread package alone for the terminal and the launcher button, or the “all in one” archive that holds piperread and piperread-gui and installs both with a single command. The two packages stay separate for anyone who only wants the terminal. The first time the interface starts without a voice, it offers to download one; nothing is downloaded until you click “Download”, and the window is also reachable from the settings. piperread --list-voices now begins with the exact command that installs the recommended voice for your language, and piperread --diagnose tells you whether the interface is installed and where to get it. The installation documentation is rewritten in English, French, German and Spanish: each chapter opens with “Do you want the interface?” and gives one numbered sequence of steps for each answer. The German and Spanish texts have not been reviewed by a native speaker. Upgrading from v0.3.3-beta needs no manual step.
v0.3.3-beta — 24 September 2026
The first beta. Your voices can no longer be hidden by a leftover folder: when PiperRead ran from a clone of the repository, a voices folder next to read.sh replaced ~/.local/share/piperread/voices/ entirely, and every voice installed there disappeared from the launcher and from the interface’s settings, without any message. Both folders are now read; when both hold a voice of the same name, the one next to read.sh is used, and --diagnose names the folder of the voice in use. The Windows interface still reads only the voices folder next to piperread-gui.exe. The README and the How it works page now show the interface icon with its menu open, on KDE Plasma, GNOME and Windows 11.
v0.3.2-alpha — 24 September 2026
A fix for the interface, which could freeze on Linux. After a right-click on the tray icon, the menu opened only after a long delay and none of its entries did anything, with no way to recover except killing the program. The clipboard tools are now given two seconds, after which the interface moves on, and a new “Restart” entry in the menu closes the interface and starts a fresh one, with the voice engine, without starting to read. The documentation now gives the command that starts the interface.
v0.3.1-alpha — 24 September 2026
A fix for the interface. A Markdown horizontal rule (---) or a line made only of punctuation stopped the reading halfway, with an error notification, and the rest of the text was never read. Such lines are now dropped before reading, and horizontal rules are removed with the rest of the Markdown markup, in the core too, which no longer reads *** aloud.
v0.3.0-alpha — 24 September 2026
The release that adds an interface. piperread-gui is a separate, optional package (Debian, Ubuntu, Fedora, openSUSE Leap and Arch), and a self-contained folder for Windows: a tray icon that starts with your session, keeps the voice loaded and prepares the next sentence while one plays. It offers pause, resume, sentence-by-sentence navigation, a settings dialog and commands to bind to keyboard shortcuts. Two durations were measured on one machine and are published on the How it works page: 1.31 to 1.34 s from the launcher to the first sound, 168 ms from a click to the first sound with the interface. The usage events of the inference library that Piper pulls in (onnxruntime, Microsoft) are now turned off by default; telemetry=on turns them back on. The desktop launcher of a clone uses the right icon and path.
v0.2.0-alpha — 22 September 2026
Settings can now be given by command-line option, environment variable or configuration file: speed, voice and language. The default speed is the natural pace of the voice, where earlier versions read 1.25 times faster; speed=1.25 in the configuration file restores it. New options --pause and --resume, and --stop (like Ctrl-C) ends the whole reading. New options --version, --diagnose and --help, and --list-voices and --download-voice: no voice is installed without a request. Messages in English, French, German and Spanish. Packages for Debian and Ubuntu, Fedora, openSUSE Leap and Arch, a user manual in four languages, and an application icon.
v0.1.2-alpha — 20 September 2026
A fixes-and-cleanup release. Voice selection now works correctly when PiperRead is installed with the English, German or Spanish voice already present; the Markdown-cleanup step no longer mangles plain text that happens to look like Markdown, such as a variable name (ma_variable) or an expression (2*3); the missing libnotify-bin dependency was added to the install instructions; the engine credit was corrected to point to its actual upstream project, OHF-voice/piper1-gpl (GPL-3.0-or-later); and the site stopped sending any request to a third-party domain.
v0.1.1-alpha — 13 January 2026
Added sanitization of Markdown and image syntax so formatting characters are no longer read aloud as words, and corrected the voice-model setup instructions.
v0.1-alpha — 11 January 2026
The first published release: the core orchestrator script for Wayland and X11, desktop integration, documentation in English, French, German and Spanish, and the MIT license.