Changelog

What's
new.

Every release we ship, in chronological order. Notes are written when each version goes out, and they're the same notes shipped inside the desktop app.

  1. October 2, 2026v1.2.4
    LinuxmacOSWindows
    • Removed 5 second audio requirement for transcription
  2. September 30, 2026v1.2.3
    LinuxmacOS

    Fixed attempting to write at cursor silent failure, on as its not possible on certain screens at the moment

  3. September 30, 2026v1.2.2
    macOS

    Fix issue where permissions wouldn't go away when enabled

  4. September 30, 2026v1.2.1
    macOS

    Fixed permission issue on MacOS

  5. September 18, 2026v1.2.0
    LinuxmacOSWindows

    Custom dictionary. Terms are applied to every transcription by the server, so they work the same on every device. The desktop and mobile apps get a Terms page (from the menu on desktop, the Terms button on the mobile home screen) to add a term with its spoken forms, edit it, turn it off and remove it, with a search box for long lists. Every change goes to the server first; the local copy is refreshed with an ETag so an unchanged list costs nothing. Settings links to the page. Project terms. The desktop app reads the names of the project folder chosen on the Terms page: the manifest name and dependencies, folder and file names two levels deep, and a .voca-terms file (spoken form = written form, one per line). The list is filtered against a stop-word list, capped at 300 terms and two seconds, cached per project for 24 hours and refreshed in the background when a recording starts with a stale cache, and sent with the audio only. File contents are never read. The first use in a project shows a notice with the term count. A switch in the panel turns the scan off. Deep scan (opt-in). Reads the identifiers in the 50 source files git history shows are edited most, keeps the names that repeat across them and are not English words or keywords (product names, internal services, acronyms, and compounds such as UsersBranch or NotFoundException), spells them the way the code does, and gives camel-case names a spoken form. Ordinary words the speech model already spells (organisation, payload, Support) and accessor names (getUserById) are dropped using a bundled English word list (SCOWL). String literals, comments, secrets and generated files are never read. Off by default; runs in the background for up to 30 seconds with a progress bar in the panel. "Add to terms" in the panel promotes any project term to the user's dictionary.

  6. September 13, 2026v1.1.1
    Windows

    No notes for this release.

  7. May 7, 2026v1.0.1
    LinuxmacOSWindows

    Continue with Google on the desktop sign-in screen. In-app update banner appears when a newer desktop build is available, so you don't have to check manually. Desktop now registers itself as a single instance and as the handler for voca:// links on launch, sign-in callbacks open the running app instead of spawning a duplicate.

  8. April 24, 2026v0.1.2
    LinuxmacOSWindows

    Whats New: 1. Added Team Support, you can now invite team members to Voca 2. Per-member usage tracking + warnings 3. 2FA sign in on desktop apps 4. Easier desktop app navigation experience

  9. April 24, 2026v0.1.1
    LinuxmacOSWindows

    1. Usage display now shows free minutes remaining alongside credit balance for PAYG accounts

  10. April 23, 2026v0.1.0
    LinuxmacOSWindows

    Initial Release