What's new

Hush changelog

What changed in each release of Hush. Pulled straight from GitHub Releases.

v1.1.0LatestOctober 16, 2026

Hush v1.1.0

Faster to control, quieter while you talk, and simpler to set up.

New in 1.1

Double-tap for hands-free. Double-tap your shortcut to keep recording without holding the key, even in Hold mode. Tap again to finish.
Paste last transcript. Press Ctrl+Cmd+V (or your own shortcut) to paste your most recent dictation again. It's also in the menu bar, along with "Copy last transcript".
Your music pauses while you dictate. Hush pauses whatever is playing and resumes it when you're done, and only if it paused it. Prefer something else? Choose Lower volume, Mute all sound or Do nothing under Settings → General → While dictating.
Simpler settings. Everyday options come first; power-user controls (paste method, typing tool, word-correction and more) live in collapsed Advanced groups. Extra AI providers sit behind "More providers".
Clearer model choice. Fast (Parakeet V3, recommended), Live preview (Nemotron Streaming, shows your words as you speak) and Most accurate (Granite Speech 4.1, for Macs with 16 GB or more).
Polish everywhere. Every screen now uses the same design system as hush.software, in light and dark mode.

Upgrading

Install over 1.0. The first launch after installing spends up to a minute preparing the speech engine for your Mac's GPU (once; after that it starts in about a second). Your settings carry over; if you had "Mute while recording" on, it becomes "Mute all sound". On first launch, go to System Settings → Privacy & Security → Open Anyway, and remove and re-add Hush under Accessibility.

What's next

Edit selected text by voice, context from the text around your cursor, styles per website, and a dictionary that learns from your corrections. See the roadmap.

Requires a Mac with Apple Silicon (M1 or later). Free and open source: https://github.com/DeepanshuPal/hush

Download Hush_1.1.0_aarch64.dmg (12.7 MB)View on GitHub

v1.0.0October 16, 2026

Hush v1.0.0

Hush is now free forever and open source under the MIT license. No trial, no license key, no account.

New in 1.0

Faster speech recognition. A new engine runs models on your Mac's GPU with Metal, about twice as fast as 0.8 on an M1 and much faster on newer chips. The model loads and warms up when Hush starts, so your first dictation is as quick as the rest.
Hold or tap. Hold Fn to talk, or tap it once to go hands-free and tap again when you're done.
Live preview. See your words as you speak.
AI clean-up, rebuilt. One system instead of three: choose a provider (Groq, OpenAI, Anthropic, OpenRouter, Cerebras, a custom endpoint, or Apple Intelligence on supported Macs), and a level. It removes fillers, fixes punctuation and keeps only what you meant after "no wait". It never answers what you dictate, and if the provider is slow, Hush pastes your raw text after 4 seconds.
Per-app styles. Default, Message, Email, Code or Raw, for any app you choose.
Voice snippets that actually fire. Matching ignores capitalization and punctuation, triggers work inside a sentence, and snippets can be edited.
Filler-word removal without AI, in many languages.
More models. Parakeet V3 (recommended), Whisper, Moonshine streaming, Canary and more, with every download verified against a checksum.
More reliable. Better paste, recovery when a microphone disconnects, a warning when another app blocks keyboard input, and recordings kept even when transcription fails.
New look. A cream and burgundy design system with a proper dark mode.

Upgrading from 0.8

Install over the old version. Your settings, history and snippets carry over, and your AI Polish key and per-app tones move to the new AI clean-up. Hush will ask you to download the new version of your speech model once.

Because Hush builds aren't notarized by Apple yet, macOS asks you to confirm the first time: go to System Settings → Privacy & Security and click Open Anyway. You may also need to remove and re-add Hush under Accessibility.