Capture system audio from whatever device the meeting app uses

[ComputelessComputer]

Meeting apps pick their own speaker, so recording the system default
endpoint went silent whenever Zoom or Teams played through something
else. macOS already taps all processes; this brings the other paths
and the AEC decision in line without asking the user to pick speakers.

Windows: activate WASAPI process loopback excluding our own process
tree, which captures every other app regardless of endpoint (Windows
11). Older builds fall back to endpoint loopback on the device another
process is actively rendering to, preferring the default.

Linux: resolve the PipeWire target / PulseAudio monitor from the sink
that has uncorked foreign sink inputs, preferring the default sink.

macOS: decide AEC and mic isolation from every output device that is
running, not just the default, so a headphone default no longer hides
meeting audio leaking out of the built-in speakers.

---

Note

Low Risk
String-catalog cleanup only; no runtime logic in the shown diff.

Overview
Removes obsolete speaker-picker copy from desktop i18n catalogs now that General audio settings no longer exposes a Speakers output selector (microphone, retention, and Remember speakers remain).

Across locale messages.po and compiled messages.ts files, Lingui drops the audio-settings strings "Speakers" and "Choose the speakers that play other participants so Anarlog records them." This keeps translations in sync with the UI change tied to the PR goal: capture meeting audio from whatever output the conferencing app uses instead of asking users to pick speakers.

^{Reviewed by Cursor Bugbot for commit e9b8df7. Bugbot is set up for automated code reviews on this repo. Configure here.}

[netlify]

✅ Deploy Preview for anarlog canceled.

Name	Link
🔨 Latest commit	e9b8df7
🔍 Latest deploy log	https://app.netlify.com/projects/anarlog/deploys/6a9958fbfee9cd0009a8a23d

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit f6a0796. Configure here.}

[cursor]

Routing poll restarts source in a loop

High Severity

OutputRoutingTracker samples headphone_only_output() in DeviceChangeWatcher::spawn before play_silence() marks the default output running and before capture_settings() opens streams. When the default is speakers and headphones are already playing, that first sample is true, later polls see false, and the source restarts. device_change does not consume the restart budget, so the cycle repeats every few seconds for the rest of the session.

Additional Locations (2)

• crates/listener-core/src/actors/source/mod.rs#L135-L147
• crates/listener-core/src/actors/source/mod.rs#L215-L218

 

^{Reviewed by Cursor Bugbot for commit f6a0796. Configure here.}

[ComputelessComputer]

Fixed in e9b8df7. Two causes: on macOS running_output_devices used DeviceIsRunningSomewhere, which counted our own silence stream on the default output (Linux/Windows already skip self_pid), so a speakers default with a meeting app on headphones flipped the verdict as soon as silence started playing. It now reads Core Audio process objects and skips our pid. The watcher is also spawned after the streams open and seeded with the verdict they opened with, so there is no second sample to disagree with.