Skip to content

SoundOff v0.2.0

Pre-release
Pre-release

Choose a tag to compare

@github-actions github-actions released this 04 Oct 16:19
· 8 commits to main since this release

SoundOff turns a recording into an editable transcript on your own computer. No account, no upload, no hosted
inference. This is a prerelease: read what each download can actually do.

New in 0.2.0

  • Two ways in. The start screen offers Transcribe a file and Record side by side, with the recorder
    right in the Record tile. Recent projects open with one click and say when you last opened them.
  • The next step where you are looking. A project with audio and no transcript offers Transcribe in the
    middle of the window, with its progress and Cancel in the same place.
  • The transcript on the waveform. Each paragraph is a band in its speaker's colour, the one being spoken is
    lit, and there is a time ruler and a readout of the time under the pointer. Double-click to play from a point;
    Shift+wheel or a sideways swipe slides along the recording.
  • Recording you cannot lose track of. A red indicator in the top bar shows a running take's clock and Stop,
    even with the side panel hidden.
  • Moving around. Alt+Up and Alt+Down step between paragraphs; double-clicking a paragraph's timestamp plays
    from it. Dragging a file over the window shows where it will go.
  • Quieter chrome. The top bar only shows what applies to what is open, and a dot beside the project's name
    marks unsaved changes.

Which download

Download What it can do
SoundOff-0.2.0-win-x64.zip Everything: import, record, transcribe, play back with the spoken word lit inside the text, correct, export.
SoundOff-0.2.0-osx-arm64.zip, SoundOff-0.2.0-osx-x64.zip, SoundOff-0.2.0-linux-x64.tar.gz Import, edit, export, bundles. No playback and no recording — those adapters are Windows-only and the app says so where their controls are.

Only the Windows build has been run. The others are built from the same source by the same workflow and
have not been launched, on any machine, by anyone. Treat them as something to try, not something to rely on.

What every build needs

  • ffmpeg and ffprobe on PATH (or in SOUNDOFF_FFMPEG_DIR). They identify imported media, decode audio for
    recognition and build playback proxies. Nothing that touches media works without them.
  • Transcription needs the private Python runtime, installed once with python scripts/setup_runtime.py, and
    then a one-time model-pack download of about 2 GB from inside the app. See the README. The runtime's paths are
    correct on macOS and Linux, but transcription has never been run there.
  • Nothing else: the app is self-contained, and New demo project… starts its fixture worker through the
    worker's own launcher, so it does not need a .NET install either.

macOS

The macOS builds are unsigned and not notarized, and they are a plain executable rather than a .app
bundle. Gatekeeper will refuse to open them until you clear the quarantine attribute yourself, and they launch
from a terminal:

xattr -dr com.apple.quarantine SoundOff-0.2.0-osx-arm64
./SoundOff-0.2.0-osx-arm64/SoundOff.Desktop

Do that only if you are comfortable running an unsigned binary you have checked the provenance of.

Linux

tar xzf SoundOff-0.2.0-linux-x64.tar.gz
./SoundOff-0.2.0-linux-x64/SoundOff.Desktop

What was verified

On Windows, on real hardware: 358 tests passed, and scripts/verify.py --clean --desktop-smoke rebuilt from
clean with a locked restore and exercised the real inference, capture and playback adapters — real WhisperX
recognition and alignment, real WASAPI loopback capture, and the real output device. The window was run and
looked at in both themes and both views. docs/VERIFICATION.md records the commands, the numbers and the limits,
including what is not covered: the new drop target and recording indicator were not seen during a real drag
or a real take, there is no screen-reader or keyboard-only pass, no performance claim, and no macOS or Linux run
at all.

The release workflow runs the same test suite on a hosted Windows runner minus the three suites that need a real
audio device or the model pack, which a hosted runner does not have.

MIT licensed. The repository's code is AI-generated; the README says so at the top.