Meet Perseus: LALAL.AI's New Groundbreaking AI for Vocal Separation

Oksana -

Vocal removal (creating an instrumental version of a song) and vocal extraction (creating an acappella) always were and still are our audience's most popular demands, so we decided to bring improvements targeted specifically at working with vocals—that's how Perseus AI came to be. 

What’s Under the Hood

The new LALAL.AI network uses transformer technology, a machine learning model first introduced by Google and now widely used in various applications, including large language models (LLMs) like OpenAI's ChatGPT. By adapting this technology to audio processing, Perseus becomes one of the first neural networks to leverage transformers for vocal separation.

At the core of Perseus lies a groundbreaking architecture that has never been used in our previous solutions. While the training procedures were similar to those used for the predecessor, Orion, we invested an average of 3.5 times more resources into developing Perseus.

The results speak for themselves: Perseus outperforms Orion by a remarkable one decibel across various metrics, translating to about 15% improvement in vocal extraction quality. This leap is equivalent to the game-changing upgrade we achieved when transitioning from the Rocknet neural network to the Cassiopeia network. 

How It Works

Perseus was designed specifically for vocal sound extraction, so the application of this neural network is only available for the “Vocal and Instrumental” stem. At least for now😉

In order to extract or remove vocals with Perseus, you don't need to do any additional actions – the new neural network is enabled by default.

All you need to do are these three simple steps:

1. Select the Vocal and Instrumental stem on the LALAL.AI home page.

2. Upload an audio or video track that contains vocals.

3. Let Perseus work its magic!

💡
If you want to check how Perseus compares to Orion, you can do it completely for free at the preview stage. After your track is processed, you will see stem previews and a settings bar where you can switch between neural networks.

Perseus AI is available by default not only for the classic LALAL.AI Stem Splitter but also for our Voice Cleaner, Lead & Back Vocal Splitter, and Echo & Reverb Remover.

In all services, the steps are the same – open the service in your browser, upload your track, and enjoy vocals isolated with surgical precision!

Special Perseus Bonus

To celebrate the release of Perseus, we've added 50 extra minutes to the Pro pack and 30 extra minutes to the Plus pack! 

Even if vocal extraction isn't your primary focus, this is an excellent opportunity to stock up on minutes and save in the long run. The packs with extra minutes are available for purchase from September 25 to October 7. Visit the LALAL.AI site during this period to grab some packs in the Pricing section.

Why “Perseus”?

Perseus AI continues our tradition of naming neural networks after constellations. The name “Perseus” was not chosen by chance—the fact that the new neural network works on a transformer model reminded us of the Perseus constellation, more specifically, of the myth of Perseus, which is closely associated with the themes of transformation.

It includes Medusa's monstrous transformation from a beautiful woman, the transformative powers of Medusa's head, and Perseus's own heroic transformation in overcoming the challenge. Transformation is a central motif throughout the Perseus myth, and we want our Perseus AI to bring you the joy of transforming your songs into studio-quality-level stems. 

We can't wait for you to try it out! Go to the LALAL.AI site to give Perseus AI a spin. 

Follow LALAL.AI on Instagram, Facebook, Twitter, TikTok, Reddit and YouTube to keep up with all our updates.