Feature
Two-source capture
Your microphone is You, the other party’s audio is Them, each captured on its own.
Each side has its own audio source plugin, so the record always says who spoke. You is your microphone. Them is system audio: from a virtual audio device such as BlackHole through cpal, or from ScreenCaptureKit, which needs no driver and excludes the app’s own audio.
A build with the anarlog-capture feature adds one shared capture for both sides with echo cancellation on your microphone. When its models are not installed, capture runs without echo cancellation and says so.
Every source produces the same canonical frames: 16 kHz mono 16-bit PCM in 20 ms frames. A span with no audio is filled with silence so the timeline stays continuous.