Description

After PR #355 (fix/empty-transcription-paste, commit 9b3a592), a silence detection gate was added to skip transcription when the recorded audio's peak RMS falls below a hardcoded threshold of 0.01. On some microphones at full input level, normal speech registers a peak RMS well below this threshold (e.g., 0.0060), causing all recordings to be silently discarded.

Steps to reproduce

1. Use a microphone whose signal level produces peak RMS values below 0.01 (even at max system input volume)
2. Record speech using OpenWhispr
3. Observe that transcription is skipped every time

Expected behavior

Speech should be transcribed normally.

Actual behavior

Every recording is rejected with:

[INFO][audio][renderer] Silence detected, skipping transcription {
  "peakRms": "0.0060",
  "threshold": 0.01
}

Suggested fix

Lower the SILENCE_THRESHOLD constant in src/helpers/audioManager.js from 0.01 to 0.002 (or lower). The current value is too aggressive for microphones with lower output levels.

Environment

• Device: Surface Pro for Business 11th Edition with Intel (SKU: Surface_Pro_11th_Edition_With_Intel_For_Business_2103)
• OS: Ubuntu 25.10, KDE Plasma 6.4.5
• Version: post-v1.5.5 (any build containing commit 9b3a592)