github RohitAg13/openWispr v1.2.0

latest releases: v1.4.0, v1.3.0
one month ago

On-device voice dictation. Download the build for your platform below.

Android (.apk): sideload it — enable "Install unknown apps" for your browser/files app, then open the APK. arm64 devices only. (Also rolling out via Google Play, staged.)

macOS (.dmg): the app is self-signed, not notarized, so on first launch macOS Gatekeeper will warn. Right-click the app → Open (once), or run xattr -dr com.apple.quarantine /Applications/OpenWispr.app.

What's new

  • macOS dictation is noticeably faster. The app now defaults to the Parakeet on-device engine (previously Whisper) and loads the speech model in the background at launch instead of on your first dictation — no more waiting on a cold model load for your first take of the session.
  • Android onboarding no longer blocks on the model download. You can continue setup while the Parakeet model finishes downloading in the background, instead of staring at a spinner.
  • Experimental: vocab-biased decoding for Parakeet on Android (off by default, opt-in in Settings). Nudges recognition toward your personal-dictionary terms; ships behind a flag because of a known upstream accuracy edge case on the biased decoding path.

Everything since v1.1.0

  • Point Android downloads at Google Play by @RohitAg13 in #28
  • Stop a Play hiccup from cancelling the whole release by @RohitAg13 in #29
  • Stop onboarding from blocking on the Parakeet download by @RohitAg13 in #35
  • Add experimental Parakeet vocab-bias decoding, off by default by @RohitAg13 in #32
  • macOS: default to Parakeet and warm the STT model at launch by @RohitAg13 in #37
  • Release 1.2.0: version bump by @RohitAg13 in #38

(openwispr.dev site/content updates in this range aren't listed here — they don't affect the app builds.)

Full Changelog: v1.1.0...v1.2.0

Don't miss a new openWispr release

NewReleases is sending notifications on new releases.