It seems that every month a new stem separation tool is released unto the audio world. So far, however, with the exception of the lesser-known Tracktion Waveform Pro, no DAWs currently feature stem separation. Until now!
Hot on the heels of Serato Sampler 2.0, FL Studio appears to be gearing up to offer native stem separation directly within the DAW architecture. The following teaser video which mysteriously surfaced on an FL Studio forum, previews FL Studio 21.2’s stem separation in action.
The Sorcery of Separation
If you’re not familiar with the concept of stem separation it’s a pretty mind-blowing process whereby individual elements of a song can be stripped out and separated from the original audio. Elements such as Vocals, Drums, Bass, Synths, Piano, and Guitar can all be separated with a single mouse click.
Stem Separation offers a myriad of useful creative applications. The ability to isolate a vocal for example can be an invaluable tool in the remixing and mash-up arenas. It’s also a great way to break down and analyze a recording. For example, if you wanted to learn to play a certain part. It can also prove extremely useful if you’re trying to replicate a bassline or chord progression in a track. Simply extract the stem you want and then use an audio-to-midi extraction tool to create the notation. (For more on Audio-to-midi conversion check out this blog)
As mentioned previously, there are numerous tools currently available that offer it. Here’s a list of some of the most popular.
- Serato – Serato Sample 2.0 and Serato DJ Pro
- iZotope – RX10
- Acon Digital – Acoustica 7.3
- Hit’N’Mix – RipX DeepRemix/DeepAudio
- Audionamix – XTRAX STEMS
- Omnisale – Lalal.ai
- Atomix Productions – VirtualDJ
- Deezer – Spleeter
It’s fair to say the initial reaction to FL’s stem addition has been rather euphoric. The news has led some to declare that other DAWs are going to be left dead in the water. While I think this will be a great addition to the already hugely popular FL Studio, I imagine FLs’ competitors are not going to hang around. Before long I think it’s likely that you will see the other major DAWs offering their own stem-separating derivatives.
I feel the proof however will eventually lie in the quality of the separation. And talking of quality according to FL Studio’s COO Scott Fisher, in order for this feature to operate most efficiently it will require internet connectivity. Fisher explains that by making this an online function, the process becomes “1,000 times faster” by giving users access to ‘terabyte-sized machine learning models.’
A local option will still be available but by utilizing an online server farm stems can be separated infinitely quicker. The online model will also allow for advanced Machine Learning models. In theory, this could mean that the quality of separation constantly develops and evolves!
As audio artifacts are always an issue when separating stems it will be very interesting to see how this will work in practice. It has already been stated that FL’s machine learning has been trained on contemporary music styles.
While I think the inclusion of stem separation within FL Studio represents an exciting development for DAW’s, this technology does already exist and is widely available. DAW integration will merely make for an easier workflow. And as the quality of stem separation improves due to advancements in machine learning the future certainly looks exciting.
What was once thought of as impossible i.e removing individual elements from a song has now become a reality. The uncharted territory of creative possibilities unlocked by this remarkable capability remains to be fully explored.