Skip to content

VoiceVector v0.3.2

Choose a tag to compare

@github-actions github-actions released this 29 Aug 19:09
· 41 commits to main since this release

Recording feel, especially with external audio interfaces.

  • Instant HUD. The recording overlay appears the moment the hotkey fires;
    the microphone opens in the background. USB interfaces (Focusrite Scarlett
    and friends) take about half a second to open, and that used to block the
    whole UI.
  • Warm microphone (Settings → General). Two independent options: keep the
    microphone open for 15 seconds after a recording (on by default — back-to-back
    takes start instantly) and always keep it open while VoiceVector runs (off by
    default). Either shows the OS mic-in-use indicator while the mic is open.
  • Level meter that works at any gain. Conservatively gained interfaces
    produce speech around −40 dBFS — perfectly transcribable, but the old linear
    meter barely moved. The meter is now a fixed, quantized scale (−60 dBFS
    silence → −15 dBFS full, three steps), and handles Float32/Int16/Int32 and
    multichannel input, taking the loudest channel.
  • Voice detection relative to the noise floor so silence-gap streaming
    (transcribe-while-you-pause) works on quiet inputs too.
  • VoiceVector --probe-audio [seconds] (macOS) prints the default input's
    format, per-channel levels, and device open latency, for diagnosing this
    class of problem.

Full changelog: https://github.com/Sammons/voicevector/blob/main/CHANGELOG.md