At the beginning of 2022, LALAL.AI presented Phoenix, an in-house trained neural network capable of superior vocal and instrumental stem splitting compared to all existing solutions on the market.
Although it was a huge milestone, we didn’t want to stop there. Our goal was to continue developing the possibilities of the network – make it separate the stems of musical instruments. We are happy to say that this goal is nearly reached!
Throughout the year we’ve been gradually adding the support of several instruments. Phoenix started to extract drums in April, then isolate bass in June. Now the neural network splits audio and video files into 2 more instrument stems – acoustic guitar and electric guitar!
Here is how acoustic & electric guitar stem separation with Phoenix compares to that of Cassiopeia, our previous neural network:
🎸Twice as fast processing. Nobody likes waiting and LALAL.AI users are no exception. Guitar isolation with Phoenix is lightning quick. Just a few seconds and you receive a preview.
🎸Better splitting accuracy. The powerful algorithm is capable of detecting all instruments that make up the track you uploaded. It takes into account a wide variety of characteristics specific to acoustic and electric guitars and extracts the guitar sound with exceptional precision.
🎸Cleaner sound. The extraction of any instrument from a mix of numerous sounds is never easy or perfect. Some sonic artifacts normally leak through an isolated stem. However, Phoenix produces significantly fewer weird sounds, providing separated stems in nearly studio-quality.
Head over to LALAL.AI to try the improved guitar splitting – select Acoustic Guitar or Electric Guitar, upload your file, and check the quality of results yourself!
📒Check out a step-by-step guide on how to extract a guitar track from any song or video here.
We are currently training Phoenix to extract the piano stem – this feature will be implemented in the near future. On our roadmap, there is also a brand new stem that LALAL.AI has never supported, wind instruments. What’s more, there is soon going to be a mobile version of the service.