Support

Getting set up

Tulpa runs on Apple Silicon Macs, an M1 or later. Intel Macs are not supported yet. Tulpa lives in the menu bar, so after setup its icon appears there rather than a window. If it will not open at all, check which chip your Mac has under Apple menu, About This Mac.

It needs two permissions on macOS. Microphone, so it can hear you, and Accessibility, so it can type into whatever app you are using. Both are requested during onboarding; you can review them later in System Settings under Privacy and Security.

If dictation seems to do nothing at all, Accessibility is almost always the reason. macOS blocks the typing step while transcription itself keeps working, so the app looks like it heard you and then did nothing. If no text field is selected, Tulpa copies your words to the clipboard instead and says so.

Shortcuts stopped working

Usually this is macOS Secure Input. While a password field is focused, or an app has switched on secure keyboard entry, macOS hides keystrokes from every other app, so your Tulpa shortcut never arrives. Tulpa shows a warning and names the app when it can tell which one.

To clear it, finish or close any password or login prompt, then quit and reopen the app Tulpa named (password managers can hold it after you unlock them). In Terminal or iTerm2, turn off Secure Keyboard Entry in the app's menu. If no app is named, lock your Mac and unlock it again; if the warning stays, log out and back in.

If pressing fn opens emoji or starts Apple's dictation, macOS is using the key too. Open System Settings, Keyboard, and set Press 🌐 key to Do Nothing, so fn belongs to Tulpa alone.

Words are missing or slow

The speech model accounts for most of this. Parakeet Unified EN, the default, is English only and returns text as you speak; the multilingual models transcribe once you stop speaking, so text arrives a beat later, and cleanup adds a moment more if it is on. Switch models in Tulpa's settings.

Accuracy also depends on input level. If your microphone records quietly, quiet syllables are harder for any model to resolve, so raising the input gain in System Settings often helps more than changing models.

Working offline

Dictation does not need a connection. Once a speech model has been downloaded, transcription runs entirely on your machine, so planes, trains and dead-signal basements are fine.

The network is used to download models, check for updates, sync your account if you sign in, and run the optional cleanup pass if you set up a cloud provider.

Privacy and what leaves your machine

Audio never leaves your device, and it is never saved. There is no setting that uploads it and no path in the app that sends it anywhere. History keeps text only, on your Mac.

From your dictation, the only thing that can ever be sent is text, for the optional cleanup pass, and only if you connect a cloud model with your own key. Cleanup is off by default, and cleanup with Apple Intelligence stays on your Mac.

Before you connect a cloud model, Tulpa warns you that the text you send goes straight from your Mac to that provider, under its own terms, and asks you to confirm. Security has the full picture.

Teams and admin

Tulpa Teams is coming soon and is not available yet. Request team access and we'll reach out when it opens.

Billing and trials

Solo is free for 14 days, with no card and no account needed to start dictating. After that it is $8 a month or $69 a year: sign in with Google, Apple or an email link to unlock it on your Mac. Words are unlimited, because transcription runs locally and has no per-word cost.

Billing is monthly or yearly, never per word, and we refund within 30 days if Tulpa is not for you. To cancel or ask about invoicing, write to us and we will sort it out directly.