Saykeep is push-to-talk dictation and bot-free meeting capture in one app. Transcripts and speaker labels are processed entirely on your computer — always. Summaries come from an AI endpoint you choose, your own machine by default. No bot. No cloud. No subscription.
Cloud notetakers add a “recorder” participant to every meeting — the awkward extra tile everyone notices and nobody invited.
Even the bot-free ones upload your voice, your meetings, your words to their servers, under a privacy policy you can’t check.
Ten to twenty dollars a month, per tool, forever — for software running on hardware you already own.
And there’s the do-it-yourself route: virtual audio drivers, loopback cables, scripts around a speech model. It works, if you enjoy maintaining it.
They removed the bot. We removed the cloud.
Your words land at the cursor (or on the clipboard) in whatever app you’re using — your editor, your email, your terminal, your chat. Fast, offline, and private: the audio never leaves the machine, so there’s no network round-trip to wait for and no word meter counting you down.
A custom vocabulary fixes the names your speech-to-text keeps mangling: colleagues, product names, acronyms. It applies across dictation, meeting transcripts, and summaries. It’s an accuracy feature the cloud tools mostly don’t have: your dictionary, on your machine.
Powered by Whisper-class open speech models running locally, in dozens of languages. On-device speech recognition now benchmarks at parity with the big cloud services (see the FAQ for the source).
One take. It lands wherever your cursor is.
Start a recording and Saykeep captures both sides, your mic and the audio your computer plays, straight from the operating system. No bot joins the call. No virtual-audio driver to install. It works with anything that plays audio: Zoom, Meet, Teams, Slack huddles, a webinar, a phone on speaker.
Watch the folder while the call runs: their voices land in one file, yours in another. Both sides, plainly yours.
Per-speaker labels (diarization) you can rename — turn “Speaker 1 / Speaker 2” into real names that stick to the transcript and summary.
Generated by any OpenAI-compatible LLM endpoint — you decide where that is. How the three positions work →
Saykeep records the way a person taking notes would — never as a hidden bot. The indicator stays on the whole time, no setting can turn it off, and your first meeting opens a consent notice. Once the transcript is saved, Saykeep can even delete the audio for you — after each meeting, or by age. Is it legal to record?
Speech-to-text, transcripts, and speaker labels always run on your computer — that part is not a setting. The only choice is who computes your meeting summaries. One setting. Three honest positions.
127.0.0.1). If your Mac is powerful enough to run a local model, everything stays home — even the summaries — and Wi-Fi can be off the whole time.Flip it all you like — the left column never changes.
Position one isn’t a promise — it’s testable: cut the network below and watch what stops working. Every connection the app can make is listed in NETWORK.md, and you can change this setting at any time.
Most apps say “private.” Saykeep is built so you can check. We publish NETWORK.md — an exhaustive list of every network connection the app is capable of making. Saykeep is closed-source, so instead of “read the code,” we made the network behavior auditable: the list is short, it’s complete, and you can watch the wire yourself.
Little Snitch, LuLu, a corporate proxy — take your pick. Saykeep keeps working in full, because it never needed the network.
Turn off Wi-Fi and record an hour-long meeting. The transcript and the speaker labels come out exactly the same — the simplest audit there is.
Press it and watch what stops working: nothing.
One payment. All local features. Every machine you own.
A future v2.0, a genuinely large new local capability, will be a paid upgrade at an owner discount. That’s how development stays funded without a subscription, and your v1 license keeps working forever.
It depends on where you and the other participants are — some places require all-party consent (Germany; California and other US states). Saykeep makes recording unmistakably visible so consent is possible: the indicator is always on, and your first meeting shows a consent notice. Obtaining the actual consent is up to you. See the full FAQ; this is not legal advice.
Saykeep runs Whisper-class open speech models locally. Recent published benchmarks show on-device Whisper-family systems at parity with the big cloud transcription services, and sometimes ahead of them (see the FAQ for the source). Names and jargon are where every model stumbles, which is exactly what the custom vocabulary is for; transcripts are plain text you can edit.
macOS 14.2+ on Apple Silicon; Intel Macs are not supported. A Linux build exists for advanced users (PipeWire system audio). Windows is in validation — join the waitlist on the download page.
Nothing breaks. Your license verifies offline — there is no activation server that could go away. Your recordings, transcripts, and summaries are ordinary files in your own folder, in open formats (Opus audio, JSON and Markdown transcripts) you can read without us.
“Saykeep exists because I wanted exactly this and nobody sold it: dictation and meeting notes that never leave my machine, from software I own instead of rent. I’ve used it every day for over a year. If you’re curious how it works under the hood, I write about the engineering.”