LALAL.AI 2022 Wrapped

As 2023 is drawing near, we are looking back on everything we’ve made and accomplished with awe and honor. 2022 was a great year for LALAL.AI, our team has doubled in size and the project has reached even more heights that we could imagine was possible in one year.

Before we dive into new year with renewed vigor, let’s go over the major changes, improvements and achievements of 2022.

🟑New Neural Network

2022 started out strong with the release of the game-changing stem separation technology, Phoenix. Like our previous neural networks Rocknet and Cassiopeia, Phoenix was developed in-house, but was qualitatively different from its predecessors in both speed and accuracy in splitting audio and video files into stems.

In terms of vocal separation quality, the new neural network surpassed not only our past solutions but also other vocal splitters on the market, including such giants as iZotope and Moises. You can read detailed analyses comparing LALAL.AI with the competitors here.

🟑Improved Splitting of 6 Stems

Initially, when Phoenix was first released, it performed an improved separation of only the vocal and instrumental stems. The other six stems supported by the service (drums, bass, acoustic guitar, electric guitar, piano, synthesizer) were separated with Cassiopeia.

Our task was to train Phoenix to separate all stems at the same level as vocals and instrumentals. Gradually over the course of a year we were moving towards this goal. Now all types of stems are supported by Phoenix, and the neural networks of the previous generations are no longer used.

🟑Support of 2 New Stems

Resting on our laurels is not LALAL.AI’s style, so we went ahead and made two more completely new stems available for splitting – wind and string instruments. With this addition, LALAL.AI has achieved the status of the world's first 10-stem splitter!

🟑New Voice Cleaning Service

In addition to expanding the stem extraction capabilities, we branched out and created a new service – Voice Cleaner. Leveraging our breakthrough technology, it cleans up voice recordings by elimnating background music and noise, including mic rumble, vocal plosives, and other extraneous sounds.

🟑Desktop Application

Although LALAL.AI is a web-based service that is available for use from any device that has a browser and Internet access, our users have often requested the release of dedicated applications.

The mobile version of LALAL.AI is still under development, but in 2022 we managed to develop and release the desktop application. It supports the main functionality of the web service and can be used on Windows, macOS, and Linux.

🟑Users & Usage Growth

Compared to last year, our service is a lot more popular in 2022. LALAL.AI currently has 1,200,000 active users (vs. 476,213 users in 2021), created 7,000,000 stem previews (vs. 2,443,049 previews in 2021), 2,400,000 full stem splits (vs. 917,506 splits in 2021), and processed 135,000 hours of audio (vs. 23,804 hours in 2021).

More than 4 million people visited the site in 2022, and the three most popular stems were vocals (73% of the total number of splits), drums (11%), and voice (6%).


The year of 2022 was a blast, and our team is excited for 2023. We wish all LALAL.AI users a very Happy New Year!πŸŽ„πŸŽ‰

Get ready for more amazing news and opportunities. We've got lots of plans, developments and new ideas to make your work with audio easier and better by all accounts.


Follow LALAL.AI on Instagram, Facebook, Twitter, TikTok, Reddit and YouTube to stay in the loop with all our updates.