I recorded the audio for the show I did with MrX in late 2019: “hpr2972 :: The foot of the ski slope”. I was using my Zoom H2n recorder in my car, on a small tripod placed on the dashboard. Something about this setup caused the result to be very boomy and (to me) unpleasant to listen to. This episode is about what I did for a cure, after some research.
I have also been using the ‘Truncate Silence’ effect in Audacity incorrectly in the past, and I used the opportunity to learn how to do a better job with it.
Now, I am well aware that there are some skilled and experienced Audio Engineers out there in HPR-land. I am certainly not one of these, though I quite enjoy fiddling with audio to make it sound better. I’d like to make two requests:
- If I didn’t do a good job, please tell me what I did wrong here, and how I should have done it.
- Think about doing a show (or shows) on HPR about how to deal with common audio problems. For example: how to remove a mains hum, the use of compression and normalisation.
Steps taken to clean the audio
I always do this because most of what I produce has background noise from my house, mains hum, etc. It’s a simple procedure. It might not have been necessary here, but I did it nevertheless.
See the Audacity Manual for a description of how this is done. In brief, the steps are:
- Sample a piece of the audio to get a Noise Profile. I tend to leave the recorder running for a short while before I start speaking so that there’s a suitable piece to sample.
- Select everything, tune the settings, and run the effect
- You should see a change to the waveform in Audacity. I usually see that the silences are a lot less noisy.
This is a way of reducing low-frequency noise. My feeling was that the “hollow” nature of the original audio was due to low frequencies echoing around the car, so I tried this to see if it helped, and found that it did.
As described in the Audacity Manual, this effect:
… passes frequencies above its cutoff frequency and attenuates frequencies below its cutoff frequency.
I set the cutoff frequency to 500.0 Hz, and the Roll-off (dB per octave) to 6dB. I tried 1000.0 Hz first but the result was truly awful - I assume it removed pretty much everything!
Since the previous filter had reduced the volume overall, I applied an amplification. This effect is documented in the Audacity Manual. When the Amplify dialog is first shown there is a value in the Amplification box which, if applied, will produce a new peak amplitude of 0dB. In my case this was 1.106dB and I just used that.
Actually this might not have been enough because the end result sounded fairly quiet when listening to it on the HPR feed, though it sounded fine played through Audacity.
There’s a certain art to using this effect properly. I find my sentences seem to trail off a bit when I speak, and this confuses the truncation algorithm. Listening back to my shows I often notice that the final word I’m saying before a pause is truncated. There are a lot of silences in the audio I produce by myself, but there were a lot fewer in this particular case.
However, when I applied the usual silence truncation settings to the audio, there were some quite unpleasant truncations at the start and end of words. This made me want to tune things so as to avoid this.
The Truncate Silence effect is documented in the Audacity Manual. Here are the settings I used:
- Threshold: -25dB. I tried -20 but some of the quieter beginnings of words were truncated
- Duration: 0.5 seconds
- Compress Excess Silence option selected
- Compress to: 50%
I had not used these setting before, but had chosen the Truncate Detected Silence option and had not been using the best Threshold value.
In the audio I have included a demonstration of a piece of audio taken from show 2972, in its original form, after noise reduction, after the high-pass filter had been applied, after amplification and after silence truncation. I hope you find it useful.
I thought the final audio sounded much better, and the silence truncation wasn’t messing up our speech as it had done before.
I hope these details will help others who need to process their audio!