You join a call and Sayli is already hearing both sides. Your microphone gives it your voice. System audio gives it theirs. What you have to grant for the second one depends entirely on your operating system, so this page is split three ways.
Sayli previews each system dialog before it appears, so you always know which button you are about to press.
Both halves of the call get transcribed
Grant both and the whole conversation is transcribed.
The microphone is not optional. Without it a call does not start at all: it is the first thing Sayli opens, and a blocked microphone stops the session before system audio is ever attached. With the microphone but no system audio, Sayli keeps going on your side only and says "Hearing you only" in its window. Settings, under Audio, shows whether each permission is granted.
What to grant, by platform
macOS 13 Ventura and later.
Microphone. A standard macOS consent dialog, once.
System audio. On macOS 14.2 and later this is its own permission, "System Audio Recording Only". macOS asks inline, the grant is immediate, and there is no trip to System Settings and no relaunch. On macOS 13.0 to 14.1 the same capture runs through ScreenCaptureKit instead, which macOS gates behind Screen Recording, so that is the toggle you will be asked for on those versions.
You can review both later in System Settings, under Privacy & Security.
Windows 10 version 2004 (build 19041) and later.
Microphone. Windows has no consent dialog for this. It is a privacy switch in Settings, and Windows does block desktop apps when it is off. If Sayli cannot hear you, it sends you straight to that pane.
System audio. Nothing to grant. Sayli captures the call through WASAPI loopback, which needs no permission.
64-bit, AppImage and .deb. On a Wayland desktop Sayli runs through XWayland, which keeps it on top and keeps its shortcuts working.
Nothing to grant. No operating-system permission gates the microphone or system audio on our Linux path. The call comes in through the PulseAudio monitor source, which PipeWire also serves.
One thing to install: a desktop keyring, gnome-keyring or kwallet.
Without one, Sayli refuses to persist your session and you sign in on
every launch. See What each platform delivers.
Skip it and there is no call to speak of: Sayli cannot start a session without it. Grant it now. Every permission is reviewable later in Settings.
Sayli previews each dialog before it appears
You grant one permission at a time, and each system dialog is previewed first, so you know exactly which button to press.
Two more rows are optional, and macOS asks nothing for either. Spot your meetings lets Sayli notice a call starting by which app is using your microphone; it is on unless you switch it off. Open at login starts Sayli when you sign in to your computer.
Staying off a screen share
It is on by default, so there is nothing to remember before a call. Sayli asks the operating system to keep its window out of a capture, and what happens next is the operating system's answer, not ours.
There is a switch in the main window's bar, and in the pane's footer during a call, if you ever want the opposite, for a demo or a screen recording of Sayli itself. On Linux it is disabled, because there is nothing there for it to turn off.
On Windows the window is removed from the capture outright. On macOS 26 and later it is removed too; on macOS 13 to 15 it is not confirmed for every screen-share app. On Linux it cannot be kept out at all. What each platform delivers states each case exactly, and the app reports what your own machine can do.
Change your microphone later
You can pick your microphone and run a live level test any time in Settings, under Audio. Do it before a call that matters, to confirm your input is coming through.