Vocalinux v0.18.1

A maintenance release on the stable line: this one is mostly fixes, with a couple of quiet additions you will notice only when you need them. Settings gains a Proxy page so model downloads and update checks work behind SOCKS5 and HTTP proxies, and an opt-in switch that persists transcripts to disk so a failed injection stops meaning a lost sentence. Releases now also attach native .deb and .rpm packages for Debian and Fedora.

Highlights

New features

• Settings Proxy page for model downloads and update checks behind SOCKS5 or HTTP CONNECT proxies, with optional authentication (#909, #655, thanks @hopsayer)
• Opt-in JSONL persistence of transcription history so a failed injection no longer loses the text (#907, #758, thanks @kacperpaczos)
• Staged Speech Model pickers that hold changes until Download is confirmed (#913, epic #600)
• Slimmer tray menu — Start on Login and About now live only in Settings (#905, #654, thanks @hopsayer)

Bug fixes

Hotkeys and input devices

• Fix: keyboard discovery on Wayland stopped grabbing mice, touchpads, and power buttons as hotkey sources (#902, #900, thanks @brainygamer)
• Fix: uinput capability bitmaps sized per platform word size — hotkeys now work on 32-bit ARM too (#917, #914, thanks @pylame22)
• Fix: letter case under KDE Plasma 6 Wayland (#918, #915, thanks @uncletoxa)

Text injection

• Fix: injection typing under apps that reject synthetic key events (#922, #921, thanks @jatinkrmalik)
• Fix: keyboard discovery leaked the dictation hotkey into the focused app (#919, #911, thanks @blackde5ert)

Wayland and windows

• Fix: dictation pad windows ghosted away under Wayland compositors (#912, #896, thanks @hopsayer)

Settings and models

• Fix: Initial Prompt in Settings > Advanced never accepted typed input; clicks were stolen by the enclosing row (#927)
• Fix: transcript and log exports could write a doubled .txt.txt filename (#927)
• Fix: model downloads that 404'd mid-fetch left broken state (#908, #894, thanks @hopsayer)
• Fix: Speech Model picker started a download per click; changes now stage until confirmed (#923, #916, thanks @blacxsnow)
• Fix: remote engine preset handed the wrong engine to the runner (#904, #677, thanks @hopsayer)

Installer

• Fix: source-build installs failed when the installer environment leaked into package builds (#901, #897, thanks @SilverChatte)

Packaging and CI

• Fix: native package dependency names differ per format — Debian/Ubuntu PySocks is python3-socks, Fedora is python3-pysocks (#920)
• Fix: AppImage boot matrix covers distros both older and newer than the build image (#899)
• Fix: release workflow attestation and integrity-check steps (#903, #906)
• Chore: dependency updates (#910)

Thanks

Bug reports that made this release: @hopsayer (#655, #654, #677, #894, #896), @kacperpaczos (#758), @blacxsnow (#916), @blackde5ert (#911), @brainygamer (#900), @pylame22 (#914), @uncletoxa (#915), @SilverChatte (#897), @jatinkrmalik (#921). Fixes and PRs: @jatinkrmalik, Dependabot (#910).

Install or upgrade

Recommended — one-line installer:

curl -fsSL https://raw.githubusercontent.com/VocaHQ/vocalinux/main/install.sh | bash

Native packages (new in this release): grab vocalinux_0.18.1_amd64.deb or vocalinux_0.18.1_x86_64.rpm (aarch64 builds too) from the assets below.

AppImage: Vocalinux-0.18.1-x86_64.AppImage (aarch64 build attached too). chmod +x and run.

AUR (source): yay -S vocalinux or paru -S vocalinux. The vocalinux-bin binary package tracks the previous stable.

PyPI: pipx install vocalinux.

Snap (sideload until the Store lists 0.18.x): sudo snap install --dangerous ./vocalinux_0.18.1_amd64.snap && sudo snap connect vocalinux:uinput.

Flatpak: flatpak install ./com.vocalinux.Vocalinux.flatpak from the assets below.

Verifying

Every binary asset ships SHA256SUMS plus GitHub attestations:

sha256sum -c SHA256SUMS --ignore-missing
gh attestation verify <file> --repo VocaHQ/vocalinux

Full changelog: v0.18.0...v0.18.1