Skip to content

audio-capture.md

Latest commit

 

History

History
95 lines (75 loc) · 4.64 KB

audio-capture.md

File metadata and controls

95 lines (75 loc) · 4.64 KB
Raw

Recording a video call

In New meeting, enable Microphone and System audio together. Choose your microphone in the first group and the output you listen to in the second. Each input has its own level meter. Adding an input keeps the device selected in the other group. You can also record either input on its own. Media file switches to importing an existing recording.

Microphone + system produces one local mono WAV at 48 kHz and sends that same mix to live transcription. Both inputs have equal gain with headroom for people speaking simultaneously. Pause and resume control both devices. A silent system output does not stop microphone recording. The recording metadata retains both device names and their original audio formats.

Use headphones to avoid capturing the call again acoustically through your microphone. The system input records all applications playing through the chosen output, including notification sounds. It does not isolate Teams or Meet.

Windows

Choose the headphones or speakers that Teams, Meet or the browser actually uses. Meet2Notes exposes their WASAPI loopback inputs under System audio. No virtual mixer is required. The application must have microphone access in Windows.

macOS

Microphone recording uses CoreAudio. System audio currently requires a virtual input such as BlackHole; Meet2Notes does not yet implement ScreenCaptureKit or Core Audio Taps.

  1. Install BlackHole separately.
  2. In Audio MIDI Setup, create a Multi-Output Device containing your headphones/speakers and BlackHole. Follow BlackHole's Multi-Output Device guide for clock and drift settings.
  3. Send the call's output to that Multi-Output Device so you can hear it and BlackHole receives it.
  4. Select your physical microphone and BlackHole in Meet2Notes.
  5. Allow microphone access for Meet2Notes, or the terminal/Python host running it.

Linux

Microphone recording uses the inputs exposed by PortAudio (ALSA/PulseAudio/ PipeWire). System audio needs an output monitor exposed as a PortAudio input. PulseAudio provides a monitor for each sink, but the ALSA/PortAudio configuration does not always list these monitors individually.

If a monitor already appears in System audio, select the one corresponding to the call's output. Otherwise, on systems using the ALSA PulseAudio plugin:

  1. Find the output monitor name with pactl list short sources (the PulseAudio compatibility server also supports this on PipeWire).

  2. Add a named input to ~/.asoundrc, preserving any existing configuration:

    pcm.meet2notes_monitor {
    type pulse
    device "YOUR_OUTPUT_MONITOR_NAME"
    hint {
        show on
        description "Meet2Notes system monitor"
    }
}

  3. Replace YOUR_OUTPUT_MONITOR_NAME with the actual monitor, restart Meet2Notes, and refresh the device list. Select your microphone plus Meet2Notes system monitor. Availability depends on the distribution's ALSA PulseAudio plugin; this is commonly provided by libasound2-plugins or alsa-plugins-pulseaudio.

See the PulseAudio monitor documentation and the ALSA PulseAudio plugin configuration. An unavailable system input is shown with platform-specific setup guidance; Meet2Notes does not silently record only the microphone when both were selected.

Device changes and troubleshooting

Refresh the list after connecting or configuring a device. If a selected device disappears, choose its replacement before starting. If one cannot be opened, both inputs are closed and the failed recording is removed so you can retry. During combined capture, a stopped stream or an input overflow is reported in the live capture strip. Save the available recording, reconnect the device, and start a new recording. Changing devices during a recording is not supported.

API compatibility

POST /api/capture/sessions still accepts source_id for one input. New clients can send source_ids: ["microphone-id", "system-id"] (one or two distinct IDs). Two inputs must contain exactly one system source and one microphone/interface. Do not send both request fields. Responses retain the primary source for older clients and include sources, source_levels, and capture_error.

The native APIs and mixer are covered by simulated Windows, Linux and macOS drivers in the test suite. Physical devices and permissions should also be checked on each target OS before distributing a release.