Skip to content

Releases

Whisper Dictate v1.0.4 — model download and doctor fixes

Choose a tag to compare

@github-actions github-actions released this 30 Sep 17:39

Whisper Dictate v1.0.4 — model download and doctor fixes

Two reliability fixes over v1.0.3:

  • First-use model downloads repaired. Model files are now fetched from
    Hugging Face resolve/ URLs, so the initial EN/DE model download no longer
    fails when the resolver handed back a preview page instead of the weights
    file.
  • Doctor mode works in the packaged app. The --doctor self-check now
    actually runs in the frozen Windows/macOS bundles instead of silently
    starting the normal dictation loop.

Also in this build:

  • Community exporter canonicalises paths (macOS /var symlink ancestors,
    Windows short names) — CI hardening, developer-facing.

Note for macOS users: this asset is the CI build — it is not yet signed or
notarized
(Developer ID signing + Apple notarization is a manual operator
step, see v1.0.3). Until a notarized build replaces it, right-click → Open
on first start, or xattr -cr WhisperDictate.app.

14-day free trial, then €40/year — whisperdictate.vercel.app.

Whisper Dictate v1.0.3

Choose a tag to compare

@github-actions github-actions released this 20 Sep 22:55

Whisper Dictate v1.0.3 — changeable keyboard shortcuts

Rebind everything without touching a config file: tray menu → "Hotkeys…" → click Change, press the new combination, Save. Conflicts are rejected, changes apply immediately, and they persist in config.json.

Covered shortcuts: EN dictate, DE dictate, Command Mode, paste-last, scratch-that, history window.

Also in this build:

  • PolyForm Free Trial 1.0.0 license text ships inside the bundles
  • Silent license self-heal: renewed subscriptions extend the app on next start
  • macOS build is signed and notarized (Developer ID + Apple notary, stapled)

14-day free trial, then €40/year — whisperdictate.vercel.app.

Whisper Dictate v1.0.2

Choose a tag to compare

@github-actions github-actions released this 20 Sep 21:24

Whisper Dictate v1.0.2

Ships the final license text (PolyForm Free Trial 1.0.0) inside the bundles, plus:

  • macOS build is signed and notarized (Developer ID, Apple notary service, stapled ticket) — opens without Gatekeeper warnings
  • Proper macOS .app bundle (the v1.0.1 mac archive was a raw folder)
  • License gate self-heal: a renewed subscription extends an installation silently on the next start — no re-activation needed
  • In-app broadcast channel for product notices (tray notification)
  • Activation with trial / purchase email, device-bound tokens (max 3), 14-day offline grace

€40/year for all use after the 14-day trial — buy at whisperdictate.vercel.app.

Whisper Dictate v1.0.1

Choose a tag to compare

@github-actions github-actions released this 20 Sep 18:36

Whisper Dictate v1.0.1 — licensing + activation

Licensing is now built in: start with a free 14-day trial (email only, no
card) or activate with the email you purchased with. Licenses are €40/year
for all use, per user, on up to 3 machines, with a 14-day offline grace so
traveling without internet never blocks dictation.

  • Activation dialog at startup: trial / activate / buy
  • Automated license registration through Stripe checkout
  • Background license revalidation; nothing about your dictation content is
    ever sent anywhere — only an anonymous email + device check at startup
  • macOS build is now signed and notarized (Developer ID, Apple notary
    service, stapled ticket) — it opens without Gatekeeper warnings
  • All the v1.0.0 dictation features (EN/DE offline models, auto-edits,
    Command Mode, voice shortcuts)

Downloads below; buy at whisperdictate.vercel.app.

Whisper Dictate v1.0.0

Choose a tag to compare

@github-actions github-actions released this 20 Sep 18:00

Whisper Dictate v1.0.0 — first public release

Local, offline push-to-talk dictation for Windows and macOS. Hold a hotkey, speak, release — polished text lands at your cursor in any app. Your audio never leaves your machine.

Highlights

  • Offline dictation (EN/DE) on NVIDIA Nemotron streaming models — GPU-accelerated (CUDA / Apple GPU), automatic CPU fallback
  • Auto-edits: filler-word removal, spoken punctuation ("comma", "new line", "bullet point", …), sentence casing — deterministic, no cloud
  • Backtrack: say "scratch that" mid-dictation to drop everything before it
  • Command Mode (hold Ctrl+Shift+F10): "select all", "paste", "undo that", "delete last word", …
  • Voice shortcuts: trigger phrases that expand into full text (snippets-*.txt)
  • Personal vocabulary: hotwords, corrections, gitignored *.local.txt overrides, fuzzy hotword reconciliation
  • Scratch-that hotkey, paste-last hotkey, clipboard-preserving paste, --doctor self-check, quiet CC0 sound themes

Install

  1. Download the zip for your platform below and unzip it.
  2. Run WhisperDictate.exe (Windows) or the app (macOS — grant Accessibility + Input Monitoring).
  3. On first dictation the speech model (~2.4 GB per language) downloads once from Hugging Face; everything afterwards is fully offline.

Licensing

The Windows bundle ships CPU torch and works on any x64 Windows 10/11 machine; with a CUDA GPU, transcription runs on it. The macOS bundle is built for Apple Silicon (arm64).