Music production software is now so advanced that genuine leaps forward have become few and far between, but the emergence of AI-powered ‘stem separation’ technology has definitely broken new ground.
Previously, it was considered all but impossible to split a mixed audio file into its various parts - vocals, drums, bass etc - but it can now be done at the click of a button.
With so many stem splitting options on the market, though, how do you choose which one to use? Inevitably, some will be better than others, and LALAL.AI believes it can now lay claim to having the best solution on the market.
One of the reasons for this is that LALAL.AI uses a different method of stem splitting to many of its rivals. Instead of taking the ‘mask-based’ approach - so called because this uses spectral ‘masks’ to hide the parts of the audio that you don’t want to hear - LALAL.AI’s constantly evolving neural network actually synthesizes and recreates the stems, giving cleaner results with fewer artefacts. It promises to work particularly well on vocal parts, with noticeably better quality than you’d get from a mask-based solution.
LALAL.AI has also paid special attention to the distortion and phasing effects that can be present when abruptly extracting stems in places where other sources are more prominent.
Another area in which LALAL.AI scores is in the number of stems that it can extract. Typically, you’re restricted to just drums, bass, vocals and ‘other’ instruments, but LALAL.AI enables you to dive in deeper and specifically isolate the likes of piano, synth, string and wind parts. You can also differentiate between electric and acoustic guitars.
This means that, whether you’re remixing or re-editing, you have far more flexibility, and this is increased even more by the Enhanced Processing option. Designed to reduce cross-stem leakage and improve stem quality, this is turned on by default, but you can turn it off if the results aren’t to your liking. Enhanced Processing can be applied to the vocal, Instrumental, drums, piano, acoustic guitar and electric guitar stems.
And that’s not all, because LALAL.AI also offers one or two features that you might not expect to find on a stem splitting platform. Advanced algorithms and machine learning enable you to remove echo and reverb from a vocal, if you wish, so you can output ‘dry’ audio that you can then process however you like. You can choose to do this when you upload a track: just select the ‘De-echo vocals/voice’ option from the drop-down menu.
There’s even a voice changer (currently in beta) that can make either a mixed song with vocals or a voice recording sound like one of your favourite artists.
You can try LALAL.AI for free, though you won’t be able to download your stems. A paid package will give you this option and others - check out the LALAL.AI website for more details.
