Skip to content

Voice-to-text dictation app with local (Nvidia Parakeet/Whisper) and cloud models (BYOK). Privacy-first and available cross-platform.

fix: request stereo recording to prevent mono transcription failure on Linux

Gabriel Stein1a407db0cdd0
diff --git a/src/helpers/audioManager.js b/src/helpers/audioManager.js
index 1617c9421..0974ccc30 100644
--- a/src/helpers/audioManager.js
+++ b/src/helpers/audioManager.js
@@ -241,10 +241,12 @@ registerProcessor("pcm-streaming-processor", PCMStreamingProcessor);
// AGC enabled to boost quiet/soft speech — essential for low-volume voice recognition.
// Echo cancellation and noise suppression stay off to avoid latency and speech distortion.
+ // Stereo recording required — mono WebM breaks silence detection on Linux/PipeWire (#472).
const noProcessing = {
echoCancellation: false,
noiseSuppression: false,
autoGainControl: true,
+ channelCount: 2,
};
if (preferBuiltIn) {