

Spake for Mac
Double-tap right Option and say what you think about anything on screen. Spake saves your words with a video of your screen, in one note.
Free for 14 days, then $29 once.
macOS 15 or later · Apple silicon · about 30 MB download1

Demo voices are synthesized. Titles added by hand.2




Double-tap right Option. Talk. Double-tap again. Spake records the screen in between.

Go on. Double-tap your right ⌥, or click the keys.Tap the keys.
Hold Shift to record just an area.Everything on screen, and what you say about it.

Not recording.
Your words, beside a recording of what you saw. Play it back and each word lights up as it’s said.2
Okay, this one is for Maya.
Look at that balcony.
Second floor, facing south, and it seats four, so we can actually have breakfast out there.
It's 118 a night, so four nights still comes in under the budget we agreed.
The dates work too: the 12th to the 16th of November, right after her birthday.
The lane is quiet and the castle steps are two minutes away.
Let's book it tonight before someone else does.
Scrolling, watching, shopping, planning. A deck, a design, a spreadsheet, some code. The same two taps keep what you said with what you saw.








Search finds words that were only on screen.




The video’s voice and yours, each in its own colour.2,3,4


Detected and transcribed on your Mac.2,5


Markdown, audio and video, in a folder you own.
Hand new notes to your agent →and 3 more
Your license is a key. You paste it once. There is no Spake account.
Recording, transcription, and search happen on your Mac, and work without Wi-Fi once the speech model has downloaded. Every request is listed in the note format.
$29 once. Nothing expires, nothing renews. There is no other tier.
Delete Spake and every note still opens: Markdown, WAV, MP4, in Finder. The app is the least permanent thing in the folder.
Payment is processed by Polar, which adds sales tax where it applies and appears on your receipt.
macOS 15 or later · Apple silicon · about 30 MB download · 714 MB–1.3 GB speech models on first run
Still stuck? Email support@spake.app and a person answers — the same one who wrote the app.
Your notes live in a folder you pick, one directory per note: Markdown, JSON, WAV, MP4, and WebVTT that open without Spake. Recording, transcription, screen text, speaker labels, search, and titles run on your Mac. Spake also keeps a search index on your Mac, outside that folder, and rebuilds it from those folders if you delete it. What Spake does send — update and license checks, model downloads, and usage statistics only if you turn them on — never carries a recording, a transcript, a frame, or a note. Every request is listed in the note format.
Recording works offline from the first launch. Transcription works offline once the speech model for your language has downloaded, once, in the background: about 714 MB for the 25 languages Parakeet covers, or about 1.3 GB when Spake needs both models, as it does for automatic language detection. Search and playback work offline too. Whisper, the second model (about 574 MB), detects the language and covers 75 more languages. Setup offers it, and otherwise it downloads the first time a language change or a note needs it. Two optional extras download separately: the 22 MB speaker model when you ask for it, and the 1.9 GB translation model, which asks before it downloads, unless you turn on Download additional languages when needed in Settings → Intelligence. Your notes stay on your Mac either way.
Apple silicon Macs on macOS 15 or later. The download is about 30 MB. On first launch Spake downloads the speech model for your language in the background: about 714 MB, or about 1.3 GB when it needs both models, as it does for automatic language detection. You can record right away, and notes transcribe when it lands. Title and label suggestions want macOS 26 with Apple Intelligence; nothing else does.
It labels the turns and lets you correct every label, which is the honest version of yes. Record system audio beside the microphone and each source keeps its own transcript; a model on your Mac groups the turns by voice, and you rename, reassign, or merge the speakers. Up to 30 minutes per source. See Who said what, above.
A hundred, detected on your Mac — the whole catalog, with the 25 that run on Parakeet marked, is in Any of 100 languages, above.
The inspector suggests up to three, each with the line that explains the match. Matching runs on your Mac over your own library — shared terms, labels, identifiers, and Apple’s on-device sentence embeddings — with no network and no vector database. They are suggestions, not saved links.
30 days, no questions. A reason helps and is not required. Email support@spake.app.
Yes, and there is nothing to install. A note is a folder of plain files, so Claude Code, Cursor or Codex can read the transcript and the frames like any others. When a new note is ready, Spake can also run a script you chose and hand it the note as JSON — that is the note.ready automation, and new ones start disabled until you turn them on. See For agents in the note format.
The pages the notes were made over — the recipe, the lecture, the headphones, the flat in Lisbon — are staged web pages built for the film, with licensed stock photos, and the deck, design, spreadsheet and code screens are staged too. The pill in the hero and carousel takes, the note rows and the search are captures of the app. The reader under “And it’s a note.” is the page’s rebuild of Spake’s reader, built from that note’s own files, and the clip plays that note’s recording. The voice reading it is synthesized, and so are the voices in every other note here. The pill over the recipe, the ring round Tram 28, the file list under “Plain files. Yours.”, the keys, the corners and the cards are drawn by the page.
The download is about 30 MB. On first launch Spake downloads its speech models in the background: Parakeet (about 714 MB) alone, or about 1.3 GB with Whisper (about 574 MB), which automatic detection needs. Recording never waits for them, and both work offline once downloaded.
The voices in these notes are synthesized (OpenAI text-to-speech), their titles and speaker names were typed by hand, and the pages they were made over are staged for the film. The notes, transcripts, timings and frames are Spake’s own output.
In this note the two voices are two sources: the lecture on system audio and you on the microphone, each with its own file, transcript and colour. System audio is on by default (Settings → Recording turns it off). Telling several voices apart inside one source takes a 22 MB speaker model, downloaded once when you ask for it, that runs on your Mac, up to 30 minutes per audio source. When two people talk at once, or the model isn’t sure, the turn stays unassigned. Labels are editable observations, not an accuracy guarantee.
Wear headphones when you record system audio: Spake does not cancel echo, and your speakers are closer to your microphone than you think.
Parakeet covers 25 of the 100 languages and Whisper the other 75; Whisper also detects which one you spoke, from the first 30 seconds. Model coverage, not an accuracy guarantee. Each audio source uses one language and engine per transcription. On macOS 26 with Apple Intelligence most notes get a title and labels in their own language, some through a translation model that Spake asks before downloading, unless you turn on automatic downloads in Settings → Intelligence. Menus come in 17 interface languages; every one but English was translated by a machine no speaker has checked.
Spake never sends them. A script, share destination or cloud-synced folder you choose can.