Saving a now-or-never take when noise strikes can test the best if the options are few. When replacement is not an option, new tools promise to kill the noise without compromising the things you need to keep…
Types Of Noise
The location recordist has a lot to take into account when capturing clean dialogue. Thanks to modern design, factors such as wind or handling noise can be easily taken care of before they hit the mic, but the one thing recordists will never have full control over is the location itself. Anyone pointing mics at talent on location will be all too familiar with the problems presented by aircraft or traffic noise, and other activity within earshot of the recording. Phones ringing ought not to be a problem on a proper production, but when capturing certain factual content the phone can interrupt dialogue that cannot be retaken or replaced after the event.
Minimizing Noise
While external factors might not be controllable, the recordist can improve things by working as close as possible to the subject. In a dry environment, if two recordings are made where a ringing phone is the same distance away from the subject, physics dictate that the closer recording will render the ringing quieter than the distant recording. This is because increasing recording the distance makes the two path lengths from dialogue and noise sources to the mic increasingly similar. Conversely, the closer the recordist works, the more different the dialogue and noise paths become as heard by the mic, with the former being picked up a higher level relative to the noise.
Replace Or Fix?
Where replacement is not an option, all is well if the noise happens in-between the wanted dialogue. If, however the noise happens simultaneously with the wanted signal two scenarios present themselves. The first is when the noise is at a level that can be worked with using conventional tools, helped and masked by other elements. The second us the worst-case scenario where noise is unacceptably high.
Using A Machine Learning Denoiser
For when the noise is persistent, we have intelligent denoisers. These remove noise, as opposed to a gate which opens and closes the signal (including noise) whenever the whole thing is loud enough to pass through. This is significantly better than edits or gating in a noise situation. Listen how I use VoiceGate2 from Accentize to remove 10 different ringtones from dialogue in the video.
Below are three of the starkest examples taken from the video, that best demonstrate the level of noise vs denoising artefacts:
Because the recording is stereo with dialogue and noise appearing in different parts of the field, further improvement could have been achieved by using the channel with the most dialogue in it. Furthermore, most of the artifacts appear to be bound to the noise in this stereo recording, opening up further options for improvement using mono tracks.
The recent v2 update to VoiceGate2 is significant, and gives you improved functionality over the original. As a completely reworked plugin, it will show up as VoiceGate2 and can be installed in parallel to prior VoiceGate 1.x versions. V2 features:
improved algorithm
new visualisation options
more efficient spectral focus mode
The Ultimate Noise Killer?
Thanks to intelligent audio plugins, mixers have an increasingly shorter list of unfixable problems. Being able to extract unwanted noise from dialogue goes much further than conventional edits or noise gating can, and being able to pull off moves like killing ringtone noise is a great example of what can now be achieved.