Have you ever wanted to remove background noise from a recording? Separate and process the vocals from a recording separately? Or remove the reverb from a vocal? Well, SpectraLayers, from Steinberg, may be the app for you.
SpectraLayers uses spectral analysis to edit and process audio tracks in different ways. Once the software has created the spectral analysis, it uses AI to remove background noise from a dialogue recording; it can separate the vocal from a backing-track to create an a cappella or instrumental; and it can clean up crackle from a vinyl transfer and remove the ambience from a recording.
What is spectral analysis?
Spectral analysis is a different way of viewing audio signals. The most common form of visual representation we use in DAWs and so forth views audio as a graph, with time on the x-axis and amplitude (volume) on the y-axis. A spectrogram views the audio in a slightly different way. Time is still the x-axis, but frequency is on the y-axis. Amplitude, or technically ‘density’, is shown through colour – the brighter the colour, the more dense those frequencies are in the signal.
Confused? Don't worry… it becomes more obvious once you use spectral analysis.
Analysing and processing audio
Getting audio into SpectraLayers to access the editing functionality is straightforward. I'm running the software in standalone mode and the audio I want to edit can simply be dragged-and-dropped into the main window. SpectraLayers then analyses the audio and presents the spectral analysis. (I believe that, when running SpectraLayers in conjunction with Cubase or Nuendo, the functionality can be accessed within the DAW itself, but I don't have Cubase on this machine, so can't test this.) Once the audio has been analysed and the spectral view is on screen, some processing can be applied. What you want to do will obviously depend on your audio, but I attempted several tasks.
De-mixing a song
One of the most common functions I need to apply to audio in my classroom is the creation of audio stems – the separate vocal, drums, bass and guitar tracks, and so on. We use this as a really useful tool to learn parts or to create backing-tracks for recordings. Within SpectraLayers 11, the function for this is under Modules > Unmix Song. The software then provides a list of likely instruments to separate. The sample audio I provided separated the drums, bass and vocals very effectively, but struggled with the keyboards and guitars and there was quite a lot of bleed between these parts.
However, because the software uses spectral analysis, it is possible to teach the software what audio should be in which parts. If SpectraLayers has mis-identified something, you can highlight the relevant frequencies and tell the software which stem it should be in.
Separating background noise
Using a similar set of functions, SpectraLayers can separate speech from background noise. I applied this to some audio of my father-in-law playing with my daughter in the park, and this was exceptional: the software removed the sound of the park and the speech into two separate stems.
I can see myself using this regularly. These days I provide most of my student feedback via audio recordings, which occasionally aren't as clear as needed. Anything that's unclear can quickly be run through SpectraLayers and the speech clarity returns.
For my own amusement I then took the park ambience and added it to the song I'd separated earlier, and created what sounded like a song being performed outside. Very clever.
Other functions
There are too many other functions in SpectraLayers 11 to explore them all in depth here, but I also explored the DeReverb function on a vocal track, and the noise, hiss and click reduction functions on some old vinyl recordings.
The DeReverb function did as described and left the original vocal dry and ready for alternative processing – great for adding either a longer or shorter reverb to a recording, or replacing unpleasant-sounding acoustics (such as a recording made in a classroom) and creating a dry sound.
The noise reduction on the vinyl was subtle but effective. I was applying it to a copy of Switched-on Bach by Wendy Carlos (unavailable digitally, so second-hand vinyl is the only option). Previous iterations of this function always had a destructive impact on the audio and, when applied too heavily, ended up sounding like the audio was underwater. Fortunately, because SpectraLayers is using AI to separate the audio via the spectral analysis, the impact is a cleaner recording without unwanted artefacts on the original.
Summary
SpectraLayers 11 is a very powerful tool, and by combining the spectral analysis and the power of AI/machine learning, it allows audio manipulation functionality that would otherwise be extremely difficult.
This software is a great addition to a music classroom and, although its functions are largely for technical tasks, in combination with a DAW or other creative software, the possibilities opened up to music departments are extensive.
SpectraLayers 11 Elements: £68
SpectraLayers 11 Pro: £254
