Record everything. Upload nothing. Zerds records both sides of a call, transcribes it and writes the notes — on your computer. Not on a server belonging to someone else, because there isn't one.
Windows, macOS, or LinuxNo account, no cloudWhat's included
Switch on airplane mode, start a recording, and watch both sides transcribe and a summary generate while the machine is disconnected. The models are already on your disk and the AI runs beside them. Nothing in that chain needs a connection.
| Call | When | Where to |
|---|---|---|
| Model download | The first time you use a model, then cached and run offline. Nothing of yours is sent — it is a file download, and every file is checked against a build-time checksum before it is used |
|
| Model catalogue | When you open the model list in Settings, to see which models exist and how big they are. Asks for a file listing — nothing of yours is part of the request |
|
| ONNX Runtime download | First run, if the runtime is not already present. Checked against a build-time checksum before it is used |
|
| Bug report / feedback form | Only when you click Report a bug or Send feedback. Opens a form in your browser — Zerds sends nothing to it, and attaches no recording, transcript or log |
|
| Your remote AI endpoint | Only if you configure one. Transcript text goes to the server you chose |
|
| Discord | Only during a Discord recording you start |
|
| Store ownership confirmation | On launch, on a Microsoft Store build. Windows confirms the licence you already own — no Zerds account, and no transcript, recording or filename is part of it |
|
In the order you'd actually do it: press Record, take the call, read the notes.
Your microphone and the system's own audio are captured as two separate channels on one timeline. It is loopback capture, not an integration, so it works with any app that makes sound — a video call, a browser tab, a phone on speaker — including apps that don't exist yet. You can scope it to a single application instead of the whole machine.
Both sides, done honestly.
Whisper runs on your machine and the text appears as people speak, labelled and timestamped. On a lighter machine you turn live transcription off and Zerds just captures during the call, then transcribes when you stop — no mid-call CPU spike.
Respect for your machine.
On Discord your own bot joins the voice channel as a visible participant and receives one separated track per person, so every line carries a real name by construction rather than by guesswork. Everywhere else, optional on-device speaker matching splits the Others channel into distinct speakers you can name.
Even working out who is speaking never leaves the machine.
A local model reads the finished transcript and writes a summary, the decisions, and the action items with owners. You pick a style preset or write your own prompt. It is grounded only in what was actually said.
The AI is local too — "private" isn't diluted the moment you want intelligence.
Meetings group into projects, and one living overview document stays current after each call while preserving your own edits. Ask it a question and the answer comes from your transcripts — and it says when something wasn't discussed rather than inventing it.
Your data stays yours, and stays usable.
The parts nobody buys software for, and everybody notices when they are missing.
Privacy here is the architecture, not a setting. There is no cloud to opt out of, no account to delete, and no server that could be compelled to hand over your conversations — because no server has ever held one. There has never been a copy to leak, sell, subpoena or train on.Why it is built this way
$34.99 · one-time · everything included
plus applicable taxes, shown at checkout
Zerds is in beta on all three platforms. It works and it is what I use every day, but you will find rough edges.
We do not offer a voluntary money-back guarantee. Refund rights required by law still apply. Stripe is the merchant of record and may issue refunds at its discretion to resolve disputes within 60 days of purchase. Read Stripe's refund policy.
No. Capture, transcription, summaries and storage all happen on your device, and there is no Zerds server for them to be sent to. The one hole in that sentence is deliberate and yours: if you point the AI at a remote endpoint you have configured, transcript text goes to that endpoint. Nothing else does. This is the complete list of network calls the application makes:
| Call | When | Where to |
|---|---|---|
| Model download | The first time you use a model, then cached and run offline. Nothing of yours is sent — it is a file download, and every file is checked against a build-time checksum before it is used |
|
| Model catalogue | When you open the model list in Settings, to see which models exist and how big they are. Asks for a file listing — nothing of yours is part of the request |
|
| ONNX Runtime download | First run, if the runtime is not already present. Checked against a build-time checksum before it is used |
|
| Bug report / feedback form | Only when you click Report a bug or Send feedback. Opens a form in your browser — Zerds sends nothing to it, and attaches no recording, transcript or log |
|
| Your remote AI endpoint | Only if you configure one. Transcript text goes to the server you chose |
|
| Discord | Only during a Discord recording you start |
|
| Store ownership confirmation | On launch, on a Microsoft Store build. Windows confirms the licence you already own — no Zerds account, and no transcript, recording or filename is part of it |
|
Not today. The Microsoft Store can run a timed trial without any server of mine, and I may add one; until then the site is the trial: the documentation is complete, the hardware check is free, and this page tells you what it can't do.
Yes. There is no subscription, no seat count, and no usage metering. On updates, precisely:
Every 1.x update is included. If there is ever a 2.0, it's a separate purchase — and your 1.x keeps working, offline, for as long as you have it.
Refunds are not mine to grant, so I am not going to promise you one. Whoever sold it to you owns that decision, and I would rather you read this site as the trial before you buy:
We do not offer a voluntary money-back guarantee. Refund rights required by law still apply. Stripe is the merchant of record and may issue refunds at its discretion to resolve disputes within 60 days of purchase. Read Stripe's refund policy.
One person, personal use, on every computer you own or control — macOS, Windows and Linux.
Yes, with two honest exceptions, both on first run: a transcription model downloads once, and the ONNX runtime downloads once if it isn't already present. After that it records, transcribes and summarises with the network off. Those downloads are in the list above, along with everything else.
You are responsible for using Zerds lawfully, obtaining any required consent, following applicable workplace rules, and securing your recordings, transcripts and exports. Recording and privacy requirements depend on where you and the other participants are. Zerds does not obtain consent for you; a Discord bot's joining notice is not a substitute for permission.
No server switches off, because there isn't one. Your recordings are ordinary audio files, your transcripts are a SQLite database, and everything exports to Markdown, SRT and JSON. There is no licence check that can start failing and no cloud that can shut down. A copy of Zerds you have installed keeps working.
If one of these matters to you, buy something else. It is cheaper for both of us to find out now.
Not sure yours can? Check your machine
A subscription notetaker in the cloud will beat Zerds on team collaboration and on cross-device sync. It buys that by holding your audio. That's the trade.
A small operation that ships seriously. Asking for money up front and staying anonymous would be an odd combination, so here is who you would be buying from.
$34.99 · one-time · everything included. Sold by Stripe. Tax shown at checkout.
Buy — $34.99 once