You don't pick a mode. Speak, and Quill works out whether you meant to write it, rewrite what you highlighted, or actually go do something — then does it, and asks before anything leaves your machine.
Quill always shows you the action before it runs it.
Auto listens to what you actually said — and to whether you've got text highlighted, and which apps you've connected — then routes it. The three explicit modes are still there for when you'd rather be the one deciding.
The default, and the reason Quill stays out of your way. One hotkey, one voice — no mode to remember, no shortcut to fumble. Say something that sounds like a task and it becomes one; say something that sounds like prose and it lands as text.
"Remind me to send the board deck Friday at 3."
Hold your hotkey and talk. Clean, punctuated text lands wherever your cursor is — in any app. Transcription runs on-device with Parakeet or Whisper; the optional clean-up pass uses Claude or GPT.
"Yes — the deck looks great. Two notes: fix slide four, add the Q3 numbers."
Highlight text anywhere, hold the hotkey, say what you want changed. Quill rewrites it in place and offers Accept or Undo — including inside Chrome and Electron apps, where most tools give up.
"Tighten this by twenty percent and make it more formal."
Speak a task and Quill fills in the right app and shows you the card before it fires. Reminders, Calendar, Gmail, Google Calendar and Todoist ship built in — plus anything reachable over MCP.
"Find Joe in Dex, then draft him a birthday email in Gmail."
One recording hotkey does all four. A second, optional shortcut cycles modes when you want to be explicit — and the menu bar always shows which one you're in.
Most voice tools stop at the transcript. Quill keeps going — it holds context about your work, takes standing instructions, and reaches tools that didn't exist when it shipped.
Describe a routine out loud and Quill saves it with a trigger phrase. Say the phrase later and it skips straight past the planner to the thing you wanted. You decide which routines are allowed to run without a confirmation.
"Whenever I say 'wrap the day', draft my standup note from today's meetings."
Quill quietly learns the people, projects and preferences that keep coming up in what you dictate, and feeds them back into planning. On Apple Silicon with Apple Intelligence this extraction runs on-device — the transcript never leaves your Mac.
"Send it to the usual list" — and it knows the usual list.
Five integrations ship built in. Everything else connects over the Model Context Protocol with a one-tap browser sign-in — Notion, Linear, GitHub, Dex, Gmail's official server, or a server you wrote this morning. Their tools become things you can say.
Everything you capture lands somewhere you can actually use it — a real markdown editor on both platforms, with your notes, photos and transcripts in sync.
Live markdown highlighting, a formatting toolbar, lists that continue themselves, and inline photos. Titles write themselves from the first few lines until you rename one.
Shorten by 20%, make it bullets, turn it into an email, extract the action items — or type your own instruction. Every change shows as a red/green diff with Undo and Keep, so nothing is rewritten behind your back.
Cloud Sync is off until you turn it on. Connect Google and your notes, photos and transcripts move between iPhone and Mac — deletes included, edits merged rather than clobbered.
Pick one and watch what the owl does with it.
Recorded on macOS in April 2026 — before Auto mode, the agent layer and the owl landed. A fresh capture is on the way.
Most voice apps stop at the transcript. The rows below the line are where Quill is doing something different.
| Feature | Quill | Wispr Flow | Superwhisper | Voibe |
|---|---|---|---|---|
| Dictate anywhere | ✓ | ✓ | ✓ | ✓ |
| Edit highlighted text by voice | ✓ | ✓ | ✓ | — |
| On-device transcription | ✓ | — | ✓ | ✓ |
| Mac and iPhone | ✓ | ✓ | Separate sub | — |
| Bring your own API key | ✓ | — | ✓ | — |
| Auto-detects what you meant | ✓ | — | — | — |
| Runs real actions in your apps | ✓ | — | — | — |
| Multi-step flows where step 2 uses step 1 | ✓ | — | — | — |
| Connect any MCP server | ✓ | — | — | — |
| Voice-authored routines & learned memory | ✓ | — | — | — |
| Actions queue offline, run on reconnect | ✓ | — | — | — |
Updated August 2026, from each product's public documentation. Every one of these is good at what it set out to do — we link to them so you can check for yourself.
Capture is unlimited and runs on your own hardware, so it costs us nothing to give you. Pro covers the part that does cost money — the AI we pay for on your behalf.
Everything the app does on your own machine, with your own keys.
No account required. No card, because there's nothing to pay.
The same app, plus the AI on us — no keys to manage and nothing to configure.
Pro is in private beta while we finish billing.
We haven't set a price yet. Quill is in beta and there's no billing in the app — so rather than post a number we might change, we'd rather you try it and tell us what it's worth. Free stays free either way: on-device capture doesn't cost us anything to run.
Specifics, not adjectives — including the cases where something does leave your device, because pretending otherwise would be worse than saying it.
Parakeet and Whisper run locally through Core ML. Your audio isn't uploaded, and there's no server in the path — Quill works with the network off.
On Macs with Apple Intelligence, the pass that learns your people and projects runs on-device via Apple's Foundation Models. Without it, that step uses your configured provider instead.
AI clean-up sends the text to your chosen provider — or to our proxy if you're on Pro. Cloud sync is off until you enable it. Crash reporting is off unless you turn it on. No tracking, no ad SDKs.
Free while we're in beta. Transcription runs on your Mac. The owl will be waiting in the corner of your screen.