How to improve audio quality of spoken word recording for transcription purposes

Hi all,

I am doing a research project that involves interviewing people using the Voice Memo app on my iPhone and then transcribing their interviews. For one of my interviews, I accidentally left my Bluetooth headphones connected, and the recording reverted to using the headphone mic (which I had stuck in a drawer). As a result, the audio quality is pretty poor, and I am having trouble producing an accurate transcription from the recording. The quality problems seem two-fold:

  • a pronounced hissing noise that continues throughout the recording.
  • a muffled quality to the vocals, especially for my interviewee (who was seated farther away from the headphones drawer than I was.)

I wish I could post sample audio, but because I need to maintain research confidentiality, I cannot. I tried applying noise reduction, and that reduced the hissing noise some, but not enough to make my interviewee’s words legible. The recording does not have to sound great by any means, just good enough that I can make out the words.

Does anyone have suggestions for ways to improve this audio file? Thanks in advance!

I am running Audacity 2.3.0 on a Windows 10 Enterprise machine.

The human mind has an incredible ability to recognise spoken words. If the sound quality is so bad that you can’t make out the words, then it is likely to be too far gone to make significant improvements (forget about the CSI tricks, that’s fiction). One thing that may help a little, is to filter out frequencies below about 300 Hz, and above 7000 Hz (See: https://manual.audacityteam.org/man/filter_curve.html). Doing that will remove or reduce sounds that are outside of the main frequency range of speech. If that does not help enough, you are probably out of luck.

Hi Steve,

Thank you for your suggestion. I did as you said and it resulted in a marginal improvement. I think I should be able to make a reasonable transcription from this file.