WhisprStream 1.0.1

WhisprStream 1.0.1 improves first-run guidance, recording controls, reliability, and future upgrades while keeping speech processing on-device.

Highlights

• Record a custom dictation shortcut, including modifier chords and F13–F24, while keeping Right Option as the default.
• Import and export vocabulary and Voice Shortcuts with additive, conflict-safe migration.
• Check for, authenticate, install, and relaunch future stable updates directly from Settings → About. Failed launches automatically restore the previous app.
• More reliable microphone recovery after audio-device changes and a first-dictation coach that waits until the app is actually ready.
• Stronger release integrity: pinned update/signing trust roots, signed assets, and hardened runtime/app validation.
• Experimental optional/custom models remain disabled in this public build.

Thanks to @FrankYang0610 for correcting the macOS model-cache path in #3.

Installing and upgrading

Download WhisprStream-macos-arm64.zip, extract it, move WhisprStream to Applications, and open it. WhisprStream supports Apple-silicon Macs running macOS 14 or later.

Public builds use a stable self-signed identity and are not Apple-notarized. On first launch, macOS may require System Settings → Privacy & Security → Open Anyway. Do not disable Gatekeeper.

Version 1.0.0 does not contain the new in-app updater, so existing 1.0.0 users must install 1.0.1 manually. Once 1.0.1 is installed, later stable versions can be installed in place from the app.

The separately downloaded speech runtime and models are preserved across app upgrades. Version 1.0.1 intentionally reuses the immutable 1.0.0 runtime asset.

Verification

SHA256SUMS covers the app ZIP, detached Ed25519 signature, and runtime ZIP. The app ZIP SHA-256 is:

9d5a5b4a3503f90d97722727e461d8632b65ef928bfb25c6f658c0f6035bf937

WhisprStream 1.0.0

WhisprStream 1.0.0

WhisprStream is a native Apple-silicon macOS dictation app that transcribes locally and inserts text at the cursor.

Highlights

• Private on-device transcription with MLX and Qwen3-ASR
• Hold-or-toggle global dictation shortcut
• Guided first-run engine and model installation
• Separate, verified speech-engine download that survives app updates
• Engine repair and scoped cleanup for interrupted model downloads
• English, Chinese, and smart bilingual handling for short dictation
• Voice shortcuts, vocabulary hints, spoken-number formatting, and context-aware capitalization
• Settings for models, sound, shortcuts, permissions, engine health, and updates

Requirements

• Apple-silicon Mac
• macOS 14 or later

Installation

Download WhisprStream-macos-arm64.zip, move WhisprStream to Applications, and open it. This build uses a stable self-signed code-signing identity and is not Apple-notarized, so the first launch may require System Settings → Privacy & Security → Open Anyway.

The app downloads the verified speech engine during setup, followed by the speech model you select. Audio and transcripts remain on your Mac.

This release is initially marked as a prerelease while the downloadable artifacts complete clean-Mac qualification.