Repository navigation
Releases: fisherinnovation/quill
Release list
Quill 0.3.0
An Action items page that gathers what your meetings agreed to do, your apps' own icons beside your dictations, and a measure of how long you wait for dictated text. Short dictations are no longer typed twice, and on the Mac they're quick again.
New
- Action items. A new page in the sidebar gathers the action items from every meeting's notes, so they don't stay buried there. Yours come first, most pressing at the top, then what you're waiting on others for, by person. Tick one off on the page, in the meeting's notes or on Home: it's the same item everywhere. Due dates are worked out from the day of the meeting ("by Thursday", "end of next week"), and flagged when overdue or due within three days. Your existing notes are read when you update.
- Your apps' icons beside your dictations, in Top apps on Home and in the per-app styles, in place of their first letter.
- Time to insert. How long you wait from your last word to the text appearing: the moment you keep holding the shortcut, then Quill transcribing and typing. The Dictations page shows the last week against the week before, how it splits, and 30 days of it; each dictation shows its own. It counts dictations made after you update.
- Windows: choose the graphics card that transcribes in Settings → Models. Quill now prefers a dedicated card to integrated graphics on its own, and the CPU is there for a card that misbehaves.
Fixes
- Short dictations are no longer typed twice ("Can you send it over? Can you send it over?"), on both macOS and Windows. On the Mac they're also back to about 0.2 s from letting go of the shortcut to typed text.
- Windows: dictated text lands in your text box, not in Quill's own window.
- Meetings: "← Meetings" stays in view as you scroll, and the darker band at the top of a meeting is gone.
Download
| System | File |
|---|---|
| macOS 14.6 or later, Apple Silicon | Quill_0.3.0_aarch64.dmg |
| Windows 10 or 11 (64-bit) | Quill_0.3.0_x64-setup.exe (recommended), or Quill_0.3.0_x64_en-US.msi |
New to Quill? Builds aren't code-signed yet, so the first launch needs one extra approval: see Installing.
Already using Quill? It offers this update itself, or use Settings → Updates → Check now.
Known issues are in the README, and how Quill handles your data is in SECURITY.md. Please report problems in Issues.
Full changelog: v0.2.0...v0.3.0
Quill 0.2.0
Notes while you talk, a prompt to record when a call starts, live captions as you dictate, and automatic updates from here on. On Windows, transcription no longer crashes on most PCs, and runs on the GPU.
New
Meetings
- Live notes. While you record, a panel at the edge of your screen keeps what to ask before the end, follow-ups and decisions, updated every minute, and ticks items off as the conversation deals with them. Pause and Stop are on the panel. It's hidden from screen sharing and never takes focus from your call. It uses your notes model with the last 15 minutes of the conversation; turn it off in Settings → Meetings.
- Offers to record calls (macOS). When Zoom, Teams, Webex, FaceTime, Slack, Discord or a browser has been using your microphone for about two seconds, a small prompt asks whether to record, and when the call ends, whether to stop. Quill sees only which app is using the microphone.
- Follow the write-up. After a recording, each step shows its progress, the time left and the lines as they're transcribed, and Quill tells you when the notes are ready.
- Stats for each meeting: who talked and for how long, a timeline of who spoke when, questions, the longest monologue and the words that came up most.
Dictation
- Live captions. What you've said appears above the dictation pill as you speak, and long dictations are typed sooner: a minute of speech in about half a second after you let go, not three.
- Stats for each dictation: words, pace, time saved compared with typing, and what clean-up took out.
- Dictations colour-coded by whether clean-up tidied them, you corrected them, or they were left on the clipboard, with a filter for each.
- Accuracy on your voice. With Settings → Dictation → Save dictations for testing on (off by default), review each saved dictation beside what was typed and see how many words Quill got right.
The app
- Automatic updates. Quill checks for a new version once a day and offers Restart to update; nothing installs until you click it, and never mid-recording. Updates are signed, and your permissions stay. Turn the check off in Settings → Updates.
- A new layout. Home leads with your latest notes and your usage, Meetings and Dictations have pages of their own, Settings has tabs, and the sidebar collapses to icons (⌘</kbd>).
- Light and dark themes, or follow your computer.
- Help inside the app, with search (⌘F or /) and a table of contents.
- Quill's logo in the menu bar and system tray.
Windows
- Transcription on the GPU through Vulkan (Nvidia, AMD and Intel graphics), or the CPU without one.
Fixes
- Windows: 0.1.0 crashed the first time it transcribed (a dictation or a meeting) on PCs without AVX-512, which is most of them, because whisper.cpp was built for the build machine's CPU. It's now built for any recent 64-bit PC (AVX2).
- macOS: quitting no longer ends in a crash report.
- Switches in Settings toggle when clicked.
Download
| System | File |
|---|---|
| macOS 14.6 or later, Apple Silicon | Quill_0.2.0_aarch64.dmg |
| Windows 10 or 11 (64-bit) | Quill_0.2.0_x64-setup.exe (recommended), or Quill_0.2.0_x64_en-US.msi |
New to Quill? Builds aren't code-signed yet, so the first launch needs one extra approval: see Installing.
Using 0.1.0? It can't update itself: download this version once, and Quill offers later updates itself.
Known issues are in the README, and how Quill handles your data is in SECURITY.md. Please report problems in Issues.
Full changelog: v0.1.0...v0.2.0
Quill 0.1.0
The first release of Quill: meeting notes and dictation that run on your computer.
- Meeting notes. Record any call (Zoom, Meet, Teams, a browser tab) without a bot joining. Your microphone and the call are recorded as separate tracks, the other voices are told apart, and Quill writes a transcript plus notes: a TL;DR, decisions, action items with owners, and the discussion by topic.
- Dictation. Hold ⌥ Option + Space (Alt + Space on Windows), speak, and let go: the text is typed into whatever app you're using, tidied and styled for that app. On a Mac you can use the fn key instead.
- Private by design. Audio is transcribed on your computer and never uploaded. Only text goes to the model you choose for notes and clean-up: OpenRouter, routed to providers that don't store or train on it, or a local Ollama model, so nothing leaves your machine.
Download
| System | File |
|---|---|
| macOS 14.6 or later, Apple Silicon | Quill_0.1.0_aarch64.dmg |
| Windows 10 or 11 (64-bit) | Quill_0.1.0_x64-setup.exe (recommended), or Quill_0.1.0_x64_en-US.msi |
First launch
- Approve the app once. These builds aren't code-signed yet.
- macOS: opening Quill shows that Apple couldn't check it. Click Done, then go to System Settings → Privacy & Security, and click Open Anyway next to the message about Quill.
- Windows: when SmartScreen says "Windows protected your PC", click More info, then Run anyway.
- Allow access. The welcome screen runs a short access check. Allow the microphone and, on macOS, system audio recording. For dictation on macOS, also allow Accessibility so Quill can paste for you.
- Download the speech models (about 1.7 GB, once) from the welcome screen. They come from Hugging Face and are checked against pinned hashes.
- Choose how notes are written. Paste an OpenRouter key (a one-hour meeting costs well under a cent), or choose a local Ollama model in Settings → Models.
Known issues
- Windows has had less real-world use than macOS. On some Windows machines, dictated text can be typed twice; this is being investigated. Transcription runs on the CPU on Windows, so it's slower than on a Mac.
- Builds aren't signed or notarized, and there's no automatic update yet: download new versions from this page.
- People who say only a sentence or two on a large call may be merged into someone else. Rename or re-split speakers in the transcript.
More detail is in the README, and how Quill handles your data is in SECURITY.md. Please report problems in Issues.