Does it work with Zoom, Teams, Google Meet and Discord?

Yes, and with anything else that plays sound on your PC. It records the audio devices rather than hooking into a particular app, so no plugin, bot or meeting link is needed, and the other participants see no extra attendee.

Record a Meeting on Windows with Transcript — Mic + System Audio, Who Said What

How it works

Press record before the call. Deal with the transcript afterwards.

The recorder captures sound devices, not a particular app — so there is no plugin to install, no bot to invite, no meeting link to paste, and nobody on the call sees an extra attendee. It works with Zoom, Teams, Meet, Discord, a phone call through the PC, or an interview recorded in the room.

1

Start the recording

Open Meeting recording in the Pro Console, give the session a name if you want one, and press record — or just press Ctrl+Shift+F8 and skip all of that, which is the point when the call has already started. Before the very first recording you are asked to confirm you understand what is being captured.

2

Both sides are captured

Your microphone and everything playing through your speakers or headphones are captured separately, and written to disk while recording — so an hour-long call is not sitting in memory waiting for a stop that might never come. When you stop, the two become a single stereo file: you on the left, everyone else on the right.

3

Stop — and nothing happens

That is deliberate. Stopping does not transcribe, does not upload, does not summarise. You have the audio, safely on disk, and the rest is your decision.

4

Transcribe when you want it

Pick a session from the list and start the transcription — an hour later or next week. You get a readable transcript, a subtitle file with timecodes, and the audio still there beside it.

Controls

A badge that says it is recording — and lets you stop it.

Windows shows a microphone indicator when an app listens to your mic. It shows nothing at all when the sound going to your headphones is captured. So NeuroSwitcher shows it itself, for as long as the recording runs.

Always visibleA small bar with the elapsed time and a live level meter per track, so “am I recording them, or only me?” is answered at a glance.

Move it out of the wayDrag it by the grip on its left edge. It starts at the top of the screen, which is also where browser tabs live — and it remembers wherever you put it instead.

Pause and stop on the badgeTwo buttons with the icons a player has. Everything else on it stays click-through, so the badge can never swallow a click meant for the window underneath.

Mute your microphone mid-callClick the microphone meter and your voice stops going into the file while the other side keeps recording — for the aside you would rather not have written down. Muted, the meter stops moving and draws struck through, and it looks the same if you switched the microphone off before the meeting, so you cannot be recording half a conversation without knowing.

Or never touch the mouseCtrl+Shift+F8 starts a recording and stops it again — no trip to the tray while the call is already running. Ctrl+Shift+F9 pauses, Ctrl+Shift+F10 stops. All three are configurable.

Who said what

Speaker labels without guessing.

Most tools record one mixed track and then try to work out who was talking from the sound of the voices. That is a hard problem, and it is wrong often enough to matter. We avoid it entirely: your voice and everyone else's arrive on different tracks, because they came from different devices.

Your track is your microphone

Everything you said, and nothing else. No bleed from the speakers, no other participant mistaken for you.

Their track is the system audio

Everything that played into your headphones — the far side of the call, clean, whatever app it came from.

Merged on timestamps

Each side is transcribed on its own and the two are woven back together in order, utterance by utterance, so the transcript reads like the conversation it was rather than two monologues. The stereo file is never transcribed as-is — mixing the sides is exactly what would lose the labels.

What it does not do

It does not tell two people on the far side apart — they are all "the other side". If you need individual names, add them while you read; the timecodes make that quick.

Your files

One folder per conversation, on your own disk.

A recording is not an entry in someone's database. It is a folder you can open, zip, move to another machine, or delete — self-contained, with everything about that conversation in it.

The audioOne stereo WAV — your microphone on the left, everyone else on the right. Nothing proprietary: any player opens it, and you hear who is speaking the same way you read it. The separate tracks can be kept alongside if you ever need to hear one side on its own.

The transcriptA readable text file with timecodes and speaker labels. This is the one you actually open.

SubtitlesAn SRT file, for anyone lining the conversation up with a video recording.

A manifestWhen it started, how long it ran, which devices were used — so a folder found in six months still explains itself.

Everything lives under %APPDATA%\NeuroSwitcher\meetings. Deleting is your action — or set a retention period and old sessions clear themselves. We cannot delete a recording for you, because we never had it.

Transcription

On your PC by default. The cloud only if you ask.

The same speech engine that powers dictation transcribes the recording, with the same choice of backend — and here the choice matters more, because an hour of a meeting is not the same as a dictated sentence.

On your PC (default)

The model runs locally. No internet, no charge, no upload — an hour of other people's voices stays where it was recorded. Slower than the cloud, and free.

Cloud, on purpose

If you pick a cloud backend for a transcription, that session's audio is uploaded to the provider to be turned into text. It is a deliberate, per-session choice, never a default and never automatic.

Faster with a GPU

If you have an NVIDIA card, an optional acceleration pack transcribes long recordings considerably faster on your own machine — the private path without the wait.

Punctuation, put back

Speech recognisers return quiet speech as an unpunctuated run of lowercase words, and mangle foreign terms spoken inside another language. Switch this on and the assistant repairs both after transcribing — punctuation, capitals, and an English term the recogniser spelled phonetically. It only ever corrects: a line it rewrote instead is thrown away and the original kept. Off by default, because it sends the text of the conversation to your AI provider.

Multilingual, like the rest

Russian, English and German transcribe well, including a call that switches language mid-sentence. Hebrew works in beta quality. More about the speech models →

Before you press record

You are recording other people. Tell them.

This is the one feature in NeuroSwitcher that captures something which is not yours. We would rather say this plainly on the product page than bury it in a settings dialog.

The law is not the same everywhereIn many countries recording a conversation is only lawful with the consent of everyone in it. Where you are, and where they are, both matter.

Platforms have their own rulesZoom, Teams and Slack impose terms of their own, on top of the law, and they are not identical to it.

Windows shows them nothingThere is no system indicator to the other participants when the audio playing on your PC is being recorded. They will not find out on their own.

We ask you onceBefore the first recording, NeuroSwitcher shows this in plain words and waits for you to confirm. After that it trusts you — the app cannot know who is on your call.

None of the above is legal advice, and it is not a substitute for it. If a recording matters legally, ask someone qualified in the places involved. See also the privacy page for exactly where the audio goes and does not go.

FAQ

Common questions about recording meetings.

Does it record the other participants, or only my microphone?

Both. Your microphone and the system audio — everything playing through your speakers or headphones, which is their voices — are captured as two separate tracks. It is not a screen recording, and it does not rely on your microphone picking up the speakers, so their side is clean even if you are on headphones.

Does the transcript say who said what?

Yes, and without any speaker-identification model: the two tracks came from two different devices, so your lines and theirs are already separate. What it cannot do is tell two people on the far side apart — everyone there is labelled as the other side of the call.

Is the recording uploaded anywhere?

No. Recording writes audio to a folder on your PC and sends nothing anywhere. Transcription is a separate action you start yourself, and the default transcription runs on your machine. The only upload is one you choose: picking a cloud backend for a particular transcription sends that session's audio to that provider.

Does it work with Zoom, Teams, Google Meet, Discord?

Yes, and with anything else that makes sound on your PC — a browser call, a desktop client, a phone call routed through the machine. It records the audio devices rather than hooking into a specific app, so there is no plugin to install, no bot to invite, and no extra participant appears in the meeting.

Is it legal to record a meeting?

It depends on where you and the other participants are. In many countries recording a conversation requires the agreement of everyone in it; Zoom, Teams and Slack also impose rules of their own. Windows gives the other side no indication that system audio is being captured, so they will not know unless you tell them — tell them. NeuroSwitcher asks you to confirm this once, before the first recording. This is not legal advice.

Where are the files and how do I delete them?

Each conversation is a folder under %APPDATA%\NeuroSwitcher\meetings: both audio tracks, the transcript as text, a subtitle file with timecodes, and a small manifest describing the session. Delete the folder and the recording is gone. You can also set a retention period so old sessions are cleared automatically.

Will a long meeting fill my disk or eat my RAM?

Audio is written to disk as it is captured, not held in memory, so a three-hour call is no more fragile than a three-minute one — and if the machine dies mid-call, what was recorded up to that point is already saved. The files are plain WAV, so an hour of a two-track recording is on the order of a few hundred megabytes; the retention setting exists precisely so they do not pile up forever.

Is meeting recording free?

It is a Pro feature, and Pro is switched on for everyone during the beta. Recording and transcribing on your own PC carry no usage cost at all — no per-minute charge, no quota. Only the optional cloud transcription draws on the same allowance as the rest of NeuroSwitcher Cloud.

Does it work on macOS?

Not yet. The recorder is Windows-only today; capturing system audio on macOS needs a different mechanism and its own permission flow, and we would rather ship it properly than half-working.

Windows or my antivirus asks for microphone access for “ks-meeting-host.exe” — is that expected?

Yes. Recording is done by a separate process, ks-meeting-host.exe, in the NeuroSwitcher install folder — not by the main application. It runs out of process on purpose: an hour of captured conversation must not be able to take the layout switcher down with it, or be taken down by it. That is why the prompt names a binary you have not seen before instead of “NeuroSwitcher”. Allowing it once is enough. Dictation does the same thing with its own process, ks-voice-host.exe.

Record the whole conversation. Transcribe it when you decide.