Skip to content

Releases: Hananel-Hazan/VibeSuperTonic

VibeSuperTonic 0.2.18

Choose a tag to compare

@github-actions github-actions released this 16 Oct 18:49
1e85f40

Fixes for the snap and the Flatpak, found the day 0.2.17 shipped.

VibeSuperTonic-0.2.18-linux-x64.tar.gz
VibeSuperTonic-0.2.18-amd64.snap           (built by CI)
VibeSuperTonic-0.2.18-x86_64.flatpak       (built by CI)
VibeSuperTonic-0.2.18-x86_64.AppImage
VibeSuperTonic-0.2.18-win.zip              (built by CI; unchanged)

0.2.17's store packages had been tested by CI and, for the snap, on a desktop.
The first test of the Flatpak on a desktop found three faults, and a closer
look at the snap found a fourth. None of them affects the tarball or the
AppImage.

The snap

  • Exporting a new file from a window the tray opened works. It failed
    with "the folder /run/user/1000/doc/… does not exist". The desktop's file
    dialog gave the file to a part of the snap that cannot see it. The window
    now hands that permission on to the part that can. Replacing an existing
    file had always worked, which is why 0.2.17's own check passed. A window
    opened from the application menu was never affected.

The Flatpak

  • The window opens on Wayland desktops. It closed straight away with
    "XOpenDisplay failed". The package asked for X11 only as a fallback, which
    Flatpak withholds whenever Wayland is present, and the window draws only
    through X11. It now asks for both: X11 for the window, Wayland so the hotkey
    can read the selection.
  • The tray icon appears. It never did: the sandbox may not claim the name
    the icon normally registers under, and the error ended the icon's setup. It
    now registers under the name the sandbox does allow. If you saw a
    VibeSuperTonic tray icon with the 0.2.17 Flatpak, it came from another
    installation.
  • Export, and reading the selection, work inside the sandbox. Checked on
    Kubuntu: a new file and an overwrite through the file dialog, and KDE grants
    the sandbox the Wayland protocol the hotkey reads the selection with.

Windows

VibeSuperTonic-0.2.18-win.zip is the Windows SAPI 5 engine, built from the same commit. Nothing in this release changes it.

VibeSuperTonic 0.2.17

Choose a tag to compare

@github-actions github-actions released this 16 Oct 15:44
ccc2da9

A snap and a Flatpak, so it can be installed from the Ubuntu App Center and from KDE Discover.

VibeSuperTonic-0.2.17-linux-x64.tar.gz     61 MB
VibeSuperTonic-0.2.17-amd64.snap           (built by CI)
VibeSuperTonic-0.2.17-x86_64.flatpak       (built by CI)
VibeSuperTonic-0.2.17-x86_64.AppImage

Two new ways to install, and the AppImage moves to third

Linux now offers four packages of the same build, in this order: the snap
(what the Ubuntu App Center installs, and what Kubuntu's Discover lists), the
Flatpak (Flathub, which Discover lists once Flatpak support is added), the
AppImage, and the tarball. The tarball is still the one the other three
are built from.

Both new packages are sandboxed. That changed three things, all invisible when
they work:

  • Models and settings live in the sandbox's own directory,
    ~/snap/vibesupertonic/common or
    ~/.var/app/io.github.hananel_hazan.VibeSuperTonic/data/vibesupertonic,
    because the program's directory is read-only there. Both survive updates.
  • An update does not claim the program moved. snapd renames the program's
    directory and $HOME on every refresh. The "re-bind your hotkeys" check now
    compares the paths that stay the same, so snap users do not get a false warning
    after each update.
  • In the Flatpak, the window, the hotkey and the screen reader module all find
    the same daemon
    , and a daemon started by a hotkey press keeps running after
    the press ends.

On Flathub, the Flatpak is built by Flathub, from source. Flathub does not
take prebuilt archives, so its build compiles this release's tag and espeak-ng
there, offline, from pinned sources, and runs the same packer and the same
checks as the tarball. It carries the .NET runtime that Flathub's SDK provides,
10.0.12, the same as the other packages, and gets .NET fixes when Flathub
updates that SDK. The listing appears once Flathub's
review is done; until then, the .flatpak file on this page installs the same
program.

Export works in the snap and the Flatpak

A sandbox cannot see the system's ffmpeg, so both packages carry a small one of
their own, built only for this: MP3, AAC and FLAC, LGPL, 1.9 MB, with its
source offer in LICENSE-FFMPEG.txt beside it. The tarball and the AppImage
still use the system's.

  • Saving through the desktop's file dialog works from a window the tray
    opened, not only from one opened from the menu. The tray's window used to
    get a file it was not allowed to write, and the export failed with "the
    folder … does not exist".
  • Export while VibeSuperTonic is reading aloud is refused, and says why
    until the reading ends.
    The message changes to "You can export now" when
    it does, so you know when to try again. Before, the refusal was a bare
    "busy (Speaking)", and in the snap it came out as an unrelated
    /proc/1/cmdline error.

Hotkeys and the screen reader: one command in a terminal

A sandbox cannot change the desktop's shortcut settings or Speech Dispatcher's
configuration, and asking the stores for that access would need a manual review
for something used once. So both packages carry sandbox-setup.sh, which you run
on the host. The Status tab shows the exact command for your install:

bash /snap/vibesupertonic/current/sandbox-setup.sh bind              # snap
bash /snap/vibesupertonic/current/sandbox-setup.sh speechd-install

It uses the same binding and Speech Dispatcher code as the tarball and the
AppImage.

For everyone on Linux: the screen reader installer is safer

These reach the tarball and the AppImage too, not only the store packages.

  • speechd-install undoes itself if it silences anything. It compares the
    voices Speech Dispatcher offered before the install with the ones it offers
    after. If any is missing, it restores the old configuration, restarts Speech
    Dispatcher, checks again and exits with an error. Before, it printed a warning
    and left the machine as it was, which could mean a screen reader with no voice.
  • It no longer waits forever on a stuck Speech Dispatcher. It waits on every
    process it asked to stop, and after five seconds it stops them by force.
  • speechd-install --remove works when our module is the only thing in the
    configuration.
    It used to exit silently at that point.

Fixed along the way to the store

Nothing below reached a stable channel, but anyone who installed from edge
before revision 5 met some of it:

  • The daemon could not start inside the snap (the sandbox refused to let it
    listen for the hotkey), there was no tray icon, and there was no sound.

  • A click on the tray icon could not open the window, at first because the window
    was started without its libraries, and then again after a logout, because a
    daemon that outlived the session still pointed at the old one. The daemon now
    takes the display from the latest program that talks to it.

  • The window says when the engine is not running. Stopped from the tray or
    a terminal, the engine went away and the open window still looked as if it
    worked. A bar across the top now says so, on every tab, with a Start it now
    button, and goes away when the engine is back.

  • On X11, the hotkey could stop working after you logged out and back in,
    with "could not open X display", if the background service had kept running
    from the earlier session. It now uses the display login the key press
    brings, as the tray's window already did.

  • After a restart, the hotkey read nothing in Wayland applications until the
    window had been opened once. The daemon looked for the desktop's Wayland
    connection inside the snap's private folder, where only the window's launcher
    puts it, and when it was not there settled on X11 for as long as it ran. It now
    makes that connection itself before it chooses. Found on Kubuntu on
    2026-09-27, after a reboot.

  • The Voices tab's Download buttons line up. Each row put its licence text
    first, and licences run from nothing to a long Creative Commons link, so every
    button sat somewhere else. The button now comes first, the id and details are
    fixed columns with the full text on hover, and a non-commercial licence says
    so in three words, with the link on hover.

  • bind no longer adds a second VibeSuperTonic to KDE's menu in the snap and
    the Flatpak. The entry that carries the hotkeys is now hidden, because the
    package has its own. It looked like a duplicate, and deleting it deleted the
    hotkeys too. sandbox-setup.sh status now says so if that has happened.

Checked on Kubuntu (KDE Plasma 6, Wayland), up to revision 10: the hotkey reads
the selection, also after a reboot before any window has opened; the tray opens
the window after a logout and login; Orca reads through the VibeSuperTonic
voice
; and a held update is applied from the prompt, with the screen-reader
voice running.

Three things to know about the snap

  • KDE reads new shortcuts only when you log in. After sandbox-setup.sh bind,
    log out and back in once.

  • You choose when an update happens. snapd holds an update back while any
    of a snap's programs runs, and the background service and the screen-reader
    voice usually are. When one is waiting, VibeSuperTonic now says so in a
    notification with Update now and Later, and the tray menu keeps an
    "Update now" item until you use it. Update now closes the window, the service
    and the screen-reader voice, and snapd installs the update straight away. The
    next hotkey press takes a few seconds longer while the voice loads, and the
    screen reader uses its default voice until you next log in, or until you run
    systemctl --user restart speech-dispatcher.service; the notification and
    the window both say so, and the window's note goes once the screen reader is
    speaking through VibeSuperTonic again. Later changes nothing; snapd installs
    the update by itself within 14 days.

    Once, to get here: the revisions already on edge do not have this, so
    the update that brings it still needs vibesupertonic.ctl shutdown,
    systemctl --user stop speech-dispatcher.service, the window closed, and
    sudo snap refresh vibesupertonic.

  • A hotkey press in the snap costs about a tenth of a second more than in the
    other packages: 108 ms per press, measured on Kubuntu, all of it snapd
    setting up the sandbox before our program starts. The tarball, the AppImage
    and the Flatpak do not pay it.

On stock Ubuntu (GNOME on Wayland)

The hotkey cannot read the selection there. GNOME does not offer the Wayland
protocol that allows it, which is true of every package, not just these. The
screen reader voices and the window work. The store listing says so.

Licence

As since 0.2.11, the Linux packages bundle espeak-ng and are GPL-3.0-or-later as
a whole. The source stays MIT.

Windows

VibeSuperTonic-0.2.17-win.zip is the Windows SAPI 5 engine, built from the same commit. Nothing in this release changes it. Requires the .NET 10 Desktop Runtime.

VibeSuperTonic 0.2.0

Choose a tag to compare

@Hananel-Hazan Hananel-Hazan released this 17 May 22:58

VibeSuperTonic 0.2.0

Portable Windows SAPI 5 TTS engine wrapping Supertone's Supertonic neural TTS.
10 English voices visible to any SAPI client (Balabolka, NVDA, Narrator, …).

Install

  1. Download VibeSuperTonic-0.2.0-win.zip below.
  2. Extract anywhere.
  3. Run VibeSuperTonic.exe once (UAC prompt — registers voice tokens).
  4. Models (~380 MB) download on first speak.

Requires .NET 10 Desktop Runtime (x64; x86 also for 32-bit SAPI clients).

See the README in the repo for full docs and limitations.

Full Changelog: v0.1.0...v0.2.0

VibeSuperTonic 0.1.0

Choose a tag to compare

@Hananel-Hazan Hananel-Hazan released this 08 May 22:50

VibeSuperTonic 0.1.0

Portable Windows SAPI 5 TTS engine wrapping Supertone's Supertonic neural TTS.
10 English voices visible to any SAPI client (Balabolka, NVDA, Narrator, …).

Install

  1. Download VibeSuperTonic-0.1.0-win.zip below.
  2. Extract anywhere.
  3. Run VibeSuperTonic.exe once (UAC prompt — registers voice tokens).
  4. Models (~380 MB) download on first speak.

Requires .NET 10 Desktop Runtime (x64; x86 also for 32-bit SAPI clients).

See the README in the repo for full docs and limitations.

Full Changelog: https://github.com/Hananel-Hazan/VibeSuperTonic/commits/v0.1.0