Frequently asked questions

Mac: Apple silicon (M1 or later) with macOS 26 Tahoe or later. Speech is recognised on the Mac's Neural Engine, which is why Intel Macs are not supported.

Windows: Windows 10 (version 2004 or later) or Windows 11, 64-bit. Speech is recognised on the PC's processor, or on its graphics card where one is available. 16 GB of memory is recommended for meetings.

Download the installer and run it. Speaku installs for your user only, so no administrator rights are needed, and it starts in the system tray next to the clock.

The installer is new and not yet widely downloaded, so Windows asks twice whether you trust it. Edge may say the file “isn't commonly downloaded”: open the … menu beside the download, choose Keep, then Show more and Keep anyway. When you run it, Windows may show “Windows protected your PC”: click More info, then Run anyway.

On first start, Speaku offers to run with administrator rights if your account has them. Windows keeps apps that run as administrator, such as some IT tools, from receiving keys and text from other apps; with the rights, dictation reaches them too. Without them Speaku works everywhere else. You can change it later in Settings › Privacy.

Put the cursor where the text should go, hold Right Option (⌥) on a Mac or Right Ctrl on Windows, speak, and let go. The text is typed there. Double-tap the key to keep talking hands-free and tap once more to insert; Esc cancels.

You can choose a different key under Settings → Dictation.

macOS only lets an app notice a key being held in another app, and type into that app, when it has Accessibility access. Speaku uses it for exactly that: the dictation key and inserting your text. Without it, dictation cannot work.

Grant it under System Settings → Privacy & Security → Accessibility. If Speaku is listed but dictation still does nothing, switch it off and on again there.

Only for calls. To put the other participants of an online meeting into the transcript, Speaku records the audio your meeting app plays — only that app, and only while you are recording a meeting. macOS asks once, the first time you record a call; Windows does not ask.

Speech models are not part of the app, so the download stays small. During setup you download a Whisper model, from about 490 MB to 1.6 GB; Parakeet (about 630 MB) can be chosen under Settings → Models. The model that tells voices apart in meetings is about 21 MB and is fetched with the first meeting. Everything is stored on your computer and loaded from there.

German and English, and switching between them. Leave the language on Automatic to mix them, or fix it to one under Settings → General.

After transcribing, Speaku can tidy the text: remove filler words, fix punctuation, and shape it for where it is going — a proper email in Mail, a list where you listed things. It uses a language model you choose in Settings → Models. Without one, the plain transcription is inserted.

Check that Speaku has Accessibility access (see above) and that the menu bar icon does not show a warning. If the text was copied instead, Speaku says so: press ⌘V to paste it. That happens when the app you were in did not accept typed text.

Yes. Open Dictionary in the sidebar and add them: how a word is heard, and how it should be written. Speaku spells them your way in dictations, in the live transcript of a meeting, in the notes and in meeting titles. A word added while a meeting is being recorded counts from that moment.

Click the Speaku icon in the menu bar (Mac) or the system tray (Windows) and choose Record meeting, or use Record a meeting on Home or the Meeting Notes page. For calls, Speaku also offers to record when Teams, Zoom, Webex or a meeting in your browser starts using the microphone.

When the meeting ends, a speaker-separation model compares the voices across the whole recording on your computer and assigns every word to a speaker. It then looks for names in what was said — “Thanks, Anna” — to name them. Click a name in the transcript to open Who spoke?: listen to a sample of each voice, name it, or merge two voices that are the same person.

On a Mac in ~/Library/Application Support/Speaku, on Windows in %LocalAppData%\Speaku: the recordings, and each meeting as a Markdown file you can open in any editor. Nothing is kept anywhere else.

Yes. On Meeting Notes, make a folder with New Folder and drag meetings onto it. A meeting whose notes you filed in Obsidian under a customer goes into the folder of the same name on its own.

Yes. Open a meeting and press Play, or click any word in the transcript to hear it from there. The word being said is marked as the recording plays.

Yes. The field at the bottom of Meeting Notes answers from your meetings — for example “what did we decide about the launch date?” — and links to the meeting it found it in. It uses the same language model as the notes.

No. Speech is recognised on your computer, for dictation and for meetings alike. Recordings stay on it.

Only text, and only if you set it up: to polish a dictation or write meeting notes, Speaku sends the transcript text to the language model you configured. If you configure one that runs on your computer, nothing leaves it at all. Model downloads and the daily update check are the only other connections. See the privacy notice.

On a Mac in your keychain, on Windows encrypted for your Windows user account. Only Speaku reads them.

It checks once a day and offers new versions itself; you can also check under Help → Check for updates or Settings → General → Updates. Click Install and Relaunch, and Speaku restarts itself in the menu bar. Every update is verified before it is installed. On Windows, Speaku downloads updates in the background and installs them the next time it starts.

Some company proxies, such as Zscaler, hold back new domains like speaku.app. The same files are served from our CDN, which they let through: download Speaku 0.2.32 for Mac from the CDN, or the Windows installer.

From version 0.2.6 on, Speaku looks for updates there itself. If yours is older and cannot find updates, install the current version from the link once; your settings and meetings are kept.

On a Mac, quit Speaku and move it from Applications to the Bin. Your recordings, transcripts and Whisper models stay in ~/Library/Application Support/Speaku, and Parakeet in ~/Library/Application Support/FluidAudio, until you delete those folders yourself.

On Windows, remove Speaku under Settings → Apps → Installed apps. Your recordings, transcripts and models stay in %LocalAppData%\Speaku until you delete that folder yourself.