As a sound engineer, gaining the knowledge and experience (in addition to spending the time ensuring your recorded signal is A++), is key, but there are times when you may have missed something or when you have had no input into that process and are receiving audio assets directly from a client. What can be done when there are problems with the audio and no way to return to the recording stage and rectify them?
A few weeks back I was introduced to iZotope’s RX6 noise reduction and audio repair software. From my first look at this industry leading software I was amazed at how authentically it is able to achieve what is does, whilst maintaining integrity of the original audio asset whether it be music or in a post production scenario.
This introduction arrived at an opportune moment for me personally as I am part of a team providing the audio assets for a short film, so was keen to learn more and deploy RX to improve our final outcomes. I began by watching a video by sound engineer/designer Evan Allen who is iZotopes lead product specialist. This video is below if you wish to delve into a breakdown of RX’s capabilities further:
The video gave me a basic idea around the navigation and key tools the application uses to achieve its results, but as we all know, programs like this with a ‘new to user’ interface have certain intricacies that one only ever learns the hard way: by trial and error (and a decent amount of googling)! I was extremely fortunate to be given a 15min one on one tutorial by my colleague Dr Duck who had gone through some audio repair exercises with the software the week before. This fast-tracked my experience and provided me knowledge that the video had not.
The RX software works better as an independently run program rather than being deployed within your DAW session so the first thing I did was do bounce the .WAV files that needed attention out from the DAW so they could be imported into RX. The content of these .WAV files was essentially location sound taken from the filming sessions by two microphones; a Sennheiser boom mic and also lapel mics on both actors. The ‘real world’ use of this repaired audio exercise was to provide clean guide tracks for the actors when they arrived for their ADR session and the audio repair was more of a courtesy than a necessity. The main issues inherent in the clips that were to be explored using RX were:
- Floor noise
- Noise from the crew (in one case somebody speaking)
- Clipping of the audio signal
- Lapel mic rustle
Basically the interface of RX is a spectrogram display. A spectrogram display is a detailed graph that displays your audio signals waveform represented not only in time and amplitude, but also represents frequencies. In RX the brighter orange the image, the higher the amplitude of that particular frequency. This allows the user to get a clear visual of where issues may lie which is incredibly helpful.
This program is a veritable labrynth of features but the basic interface (as seen in Fig.1), uses a vertical menu to the right of the image where specific functions such as De-click, Spectral repair, De-hum and corrective EQ appear, and menus below which control the playback of clips and also the selection tool bars.
The first stage of audio repair in each clip I worked on was to remove everything below 150Hz and this task was best performed with the ‘frequency selector’ (Fig 2. icon 11). This tool allows for a horizontal rectangular selection to be made from the bottom of the spectrograph up to the Hz level desired. Simply then clicking the delete key on the keyboard will remove all frequencies within that range. This is a very basic function of the software however and in no way represents its capabilities. Where this software really shines is with its ‘machine learning‘. This is the process whereby an event or specific frequency deemed to be problematic on the spectrograph is selected (I primarily used the ‘lasso and magic wand tools’ (Fig.2 icons 12 & 14)), for this purpose and then the spectral repair function is utilised. 
Its module offers several options including attenuate and replace. The user can then ask the software to ‘learn’ this frequency event and a complex algorithm is then deployed in the background to locate other areas in the clip with the same characteristics. Once the ‘process’ tab is clicked after RX has learned the selection, all other areas in the .WAV are repaired as well. This is a better demonstration of the real power of this software and how it can help you clean your audio.
Another compelling function in the software (that came in handy for me), was the ‘De-clip’ tool. This allows the user to identify an area on the waveform that has clipped and bring down the peak essentially squashing the offender to remove the audible problems associated with clipping. What i did find however, is that you must be quite wary of
how this will affect the overall signal/tone. The system has a ‘compute’ feature which will recommend a specific level of amplitude reduction which I found to be pretty accurate, however I discovered that being too aggressive with this tool greatly altered the sound and tone of the voice that caused the original peak in my .WAV.
In conclusion having only just scratched the surface of this amazing software, I can immediately see how its price tag of over $500AUD (for the standard version) can be justified. After an hour tutorial and some pertinent advice from a friend, I was able to greatly improve the quality of the audio clips I needed for the ADR session and completed my reconnaissance feeling impressed and empowered with the possibilities iZotopes RX offers. The interface is quite intuitive and it is clear that a lot of work has gone into getting the algorithms that drive the software’s decisions accurate. I will be using RX for a long time to come…
Abundant Audio 