Skip to content

Releases: RonanDavalan/PiperRead

PiperRead v0.4.1-Beta

Choose a tag to compare

@RonanDavalan RonanDavalan released this 13 Oct 21:29

PiperRead v0.4.1-beta

No function changes.

What's new

  • Packages for Fedora 44 and openSUSE Leap 16.0. Version 0.4.0-beta was built for Fedora 42 and openSUSE Leap 15.6.
  • The Windows archive now ships the license texts of its components (licenses folder, "Licenses" section of README.txt), the exact version of each component (licenses/COMPONENTS.txt) and a written offer of their sources.
  • The README (four languages) and the manual page state what each platform has been through.

Platforms

  • Debian 12 and 13, Ubuntu 22.04 and 24.04 (.deb): installation, update from 0.4.0-beta, diagnosis and removal validated in containers.
  • Fedora 44 (.rpm) and Arch Linux (.pkg.tar.zst): installation, update from 0.4.0-beta, diagnosis and removal validated in containers.
  • openSUSE Leap 16.0 (.rpm): installation, diagnosis and removal validated in containers.
  • Windows (piperread-gui-windows.zip, interface only): executable built on Windows and tested by hand on Windows 11.

A container proves the installation, the diagnosis and the removal; it does not prove the sound, the clipboard, the menu entry or the display.

Update from 0.4.0-beta: install the new packages over the old ones, no manual step.

Licenses of the bundled components

  • PiperRead: MIT.
  • Piper (speech engine): GPL-3.0-or-later. It is a separate program that PiperRead starts; on Linux it is installed by pip on your machine, on Windows it is in the archive with its license text and the address of its sources.
  • Qt for Python (PySide6, the interface): LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only, used under the LGPL; the archive ships the texts and the libraries can be replaced.
  • Voices are not included; each voice keeps its own license, shown before you choose it.

Good to know

  • Reading is 100 % local: no text leaves your machine.
  • The German and Spanish texts have not been reviewed by a native speaker; corrections are welcome as issues.

Downloads

Package version: 0.4.1~beta (Arch: 0.4.1beta).

Packages for Debian and Ubuntu (.deb), Fedora and openSUSE Leap (.rpm) and Arch (.pkg.tar.zst), the four "all in one" archives, the Windows archive, the four PDF manuals, the source tarball and its .sha256. Installation: https://piperread.davalan.fr

PiperRead v0.4.0-Beta

Choose a tag to compare

@RonanDavalan RonanDavalan released this 29 Sep 18:05

PiperRead v0.4.0-beta — two ways to install, one voice window

PiperRead now offers two clear ways to install it on every Linux distribution.
Pick the one that matches how you want to listen:

  • Terminal and button: install the piperread package. piperread reads the
    selection or the clipboard, with --pause, --resume, --stop,
    --list-voices and --diagnose, and a "Read the selection" button sits in your
    applications menu. Download: about 195 MB, installed: about 680 MB on Debian 13 (117 to 304 MB and
    329 to 792 MB depending on the distribution).
  • With the graphical interface: download the "all in one" archive for your
    distribution, which holds both packages, and install them with one command.
    The interface stays in your notification area and lets you play, pause, stop,
    move from sentence to sentence and adjust the voice and speed. Download:
    about 460 MB, installed: about 1.4 GB on Debian 13 (421 to 483 MB and 1.26 to
    1.49 GB depending on the distribution; the interface's dependencies, PySide6,
    are downloaded when its package is configured).

What changed

  • One archive per distribution family (Debian and Ubuntu, Fedora, openSUSE Leap,
    Arch) bundles piperread and piperread-gui. The two packages stay separate
    for anyone who only wants the terminal.
  • The interface offers to download a voice the first time it starts without one.
    Nothing is downloaded until you click "Download".
  • piperread --list-voices now begins with the exact command that installs the
    recommended voice for your language.
  • piperread --diagnose tells you whether the interface is installed and where to
    get it.
  • The documentation is rewritten in English, French, German and Spanish. Every
    installation chapter starts with the same question, "Do you want the
    interface?", and then gives one numbered sequence of steps for each answer.

Still true

  • No voice is bundled with any package, and none is downloaded without your
    request. Each voice keeps its own licence, shown before you choose it.
  • Reading is 100 % local: no text leaves your machine.
  • Upgrading from v0.3.3-beta needs no manual step: install the new packages over
    the old ones.
  • The German and Spanish texts (documentation, manual page, site) have not been reviewed by a native speaker; corrections are welcome as issues.
  • Windows: the interface remains a separate archive, piperread-gui-windows.zip.

Downloads

Packages for Debian and Ubuntu (.deb), Fedora and openSUSE Leap (.rpm) and
Arch (.pkg.tar.zst), the four "all in one" archives, the source tarball with its
.sha256, the four PDF manuals and the Windows archive. Installation:
https://piperread.davalan.fr

PiperRead v0.3.3-Beta

Choose a tag to compare

@RonanDavalan RonanDavalan released this 24 Sep 19:10

PiperRead v0.3.3-beta

Your voices can no longer be hidden by a leftover folder, and the README and
the website now show what the interface looks like.

PiperRead looks for voices in ~/.local/share/piperread/voices/, and, when it
runs from a clone of the repository, in a voices folder next to read.sh.
Until now, that second folder, when it existed, replaced the first one
entirely: every voice in ~/.local/share/piperread/voices/ disappeared from
the launcher and from the interface's settings, without any message.

  • Both folders are now read. When both hold a voice with the same name, the
    one next to read.sh is used. --download-voice still saves into the first
    folder that exists.
  • The settings dialog of the interface lists the voices of both folders. The
    Windows interface still reads only the voices folder next to
    piperread-gui.exe.
  • piperread --diagnose now names the folder of the voice in use.
  • The README, in its four languages, and the "Understand" page of the website
    show the interface icon with its menu open on KDE Plasma, GNOME and
    Windows 11. The README's list of menu entries now includes "Restart".
  • Packages for Debian and Ubuntu, Fedora, openSUSE Leap and Arch, and the
    Windows folder, are rebuilt for this version.

PiperRead v0.3.2-Alpha

Choose a tag to compare

@RonanDavalan RonanDavalan released this 24 Sep 16:45

PiperRead v0.3.2-alpha

A fix release 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; there was no way to recover
except killing the program. To read the mouse selection, the interface asks
an external program (wl-paste under Wayland, xsel under X11) for the
selected text, and it waited for the answer with no time limit. When the
application that owns the selection did not answer, the interface stayed
blocked.

  • The interface now gives these clipboard tools two seconds. If they do not
    answer in time, it moves on to the next one, then to the clipboard, as
    before.
  • The desktop notification command (notify-send), the only other external
    program the interface waits for, is also limited to two seconds. The
    Windows interface reads the clipboard differently and was not affected.
  • New: a "Restart" entry in the tray menu. It closes the interface and
    starts a fresh one, with the voice engine, without starting to read.
  • Documentation: piperread --help, the manual page (man piperread), the
    README and the user manual now give the command that starts the
    interface: piperread-gui when the package is installed, or
    gui/.venv/bin/python3 -m piperread_gui.app from a clone of the
    repository once the environment in gui/ is set up.
  • Packages for Debian and Ubuntu, Fedora, openSUSE Leap and Arch, and the
    Windows folder, are rebuilt for this version. The interface package now
    requires piperread 0.3.2-alpha or later.

PiperRead v0.3.1-Alpha

Choose a tag to compare

@RonanDavalan RonanDavalan released this 24 Sep 08:07

PiperRead v0.3.1-alpha

A fix release for the interface.

A Markdown horizontal rule (---) in the copied text stopped the interface
in the middle of the reading, with a "server answered 500" notification, and
nothing after the rule was read. The same happened with a line made only of
punctuation, such as a lone dash or ellipsis. The speech server cannot voice a
text without any sound to pronounce and fails on it.

  • Horizontal rules (---, ***, * * *, ___) are now removed with the rest
    of the Markdown markup, in the core and in the interface. The core no longer
    reads *** aloud as "asterisk".
  • The interface no longer sends the engine a sentence without any letter or
    digit, so the reading goes on to the end.

PiperRead v0.3.0-alpha

Choose a tag to compare

@RonanDavalan RonanDavalan released this 23 Sep 22:18

PiperRead v0.3.0-alpha

PiperRead now has a graphical interface, installed as a separate package
(piperread-gui) next to the core. The core is unchanged: the one-click
"Read selection" launcher still works on its own, with nothing running in
the background.

The interface sits in the notification area and starts with your session
(untick "Start with the session" in its settings to turn that off). The
voice stays loaded while it runs, so reading starts almost at once, and the
next sentence is prepared while the current one plays. Left-click the icon
to read the selection; click again to pause, and again to resume. The menu
adds stop, previous and next sentence, and settings (voice, speed,
language).

  • Linux: .deb, .rpm (Fedora, openSUSE) and Arch packages.
  • Windows: a standalone folder, no Python required.
  • On Linux the mouse selection is read before the clipboard, and Markdown
    markup is removed before reading, as in the core.

Privacy: the inference engine used by Piper (onnxruntime) sends usage
telemetry to Microsoft by default. PiperRead now turns it off by default, in
the core and in the interface, on Linux and on Windows. To turn it back on,
set telemetry=on in ~/.config/piperread/piperread.conf.

PiperRead v0.2.0-alpha

Choose a tag to compare

@RonanDavalan RonanDavalan released this 22 Sep 11:26

Install from a native package (.deb, .rpm for Fedora/openSUSE, Arch) — no clone needed. The Piper engine (~75 MB) installs separately via pip, never bundled in the package; no voice is bundled either.

The default reading speed changes from 1.25x natural speed to natural speed. To restore the previous behaviour, set speed=1.25 in ~/.config/piperread/piperread.conf.

Installation, manual and checksums: https://piperread.davalan.fr/download/

PiperRead v0.1.2-Alpha

Choose a tag to compare

@RonanDavalan RonanDavalan released this 20 Sep 17:39

Fixes and cleanup release. Installation and documentation: https://piperread.davalan.fr

  • read.sh loads the voice found in voices/, so the English, German and Spanish installs work.
  • Markdown cleanup no longer alters identifiers such as ma_variable or 2*3.
  • libnotify-bin added to the install command; engine credit now points to OHF-voice/piper1-gpl (GPL-3.0-or-later).
  • The site no longer sends any request to another domain.

PiperRead v0.1.1-Alpha

Choose a tag to compare

@RonanDavalan RonanDavalan released this 20 Sep 17:39

Adds Markdown and image sanitization so that syntax is not read aloud, and corrects the voice model instructions.

PiperRead v0.1-Alpha

Choose a tag to compare

@RonanDavalan RonanDavalan released this 20 Sep 17:39

First alpha release: core orchestrator script (read.sh) for Wayland and X11, desktop integration, documentation in English, French, German and Spanish, MIT license.