Headset selection and echo cancellation

Achieving studio-grade speech transcription depends fundamentally on input signal quality. While Whisper handles background hum exceptionally well, acoustic loopback from open speakers remains the primary cause of transcription errors.

The acoustic loopback challenge

When you participate in a conference call using your Mac’s internal speakers:

1. The remote participant speaks.
2. The sound emits from your laptop speakers.
3. Your built-in microphone picks up that sound with a 50–150ms acoustic room delay.
4. Whisper receives two overlapping streams of the same speech, leading to duplicate or garbled text.

Recommended hardware setups

Tier 1: Wired or Bluetooth Headphones (Best)

• Examples: AirPods Pro, Sony WH-1000XM5, wired USB headsets, or studio headphones.
• Why it works: Sound routes directly into your ears. The microphone captures only your vocal cords.
• Result: 100% channel isolation between Channel 0 (Your Mic) and Channel 1 (Call Loopback).

Tier 2: Directional USB Microphone + Speakers (Good)

• Examples: Shure MV7, Rode PodMic, Blue Yeti with cardioid pickup pattern.
• Why it works: Position the cardioid microphone pointing toward your mouth and away from desk speakers.
• Adjustment: Lower your speaker volume and calibrate mic input gain in macOS System Settings.

Tier 3: Built-in Mac Microphone and Speakers (Fair)

• Acceptable for quick 1-on-1 calls, but expect occasional duplicate phrase transcriptions when participants speak loudly.

Audio diagnostics in Oyma

Before an important executive call, test your audio signal in Settings › Meetings › Diagnostics:

1. Click Run check.
2. Speak into your microphone while playing sample audio.
3. Verify that the input level meters bounce cleanly in the green zone (-18 dB to -6 dB) without clipping into red.

Related guides

• Audio pipeline in System audio pipeline and ScreenCaptureKit.
• Permission check in Recording permissions on macOS.
• Troubleshooting audio in Audio device switching and troubleshooting.