LALAL.AI is a popular AI vocal remover that now lets you separate drums and bass stems with the Andromeda update. The update adds drums and bass to the model’s stem separation capabilities, which means you can now isolate the full rhythm section of any song in one go, along with vocals and instrumentals. There are many more improvements, so let’s go ahead and check out the new features in the LALAL.AI Andromeda model update.
LALAL.AI: What’s New in the Andromeda Update?
The headline feature is that drums and bass separation is now built into the Andromeda model. Earlier, if you wanted clean drum or bass stems, you had to use older models like Perseus, which meant two different neural networks and two different processing runs. The output also seemed inconsistent between stems.
Now, Andromeda handles vocals, instruments, drums, and bass inside a single network. Basically, all four parts come from the same AI neural network in one go. Not only that, but bass extraction has been significantly improved. LALAL.AI’s new Andromeda model captures more of the texture and detail of bass lines that used to get lost. Basically, harmonics, texture, and other details aren’t flattened during separation.
Drum isolation has improved in the Andromeda model, and it delivers cleaner and more consistent results across an entire song. Earlier, the drum stem would get weak or patchy in the middle of the track, but that is now much improved. So, drum stems are coherent from the start to the end instead of flattening randomly during the chorus.
Apart from that, processing speed is another big upgrade with the Andromeda update. It runs up to 40% faster than the previous Perseus model, which is pretty helpful if you are working through a full session or preparing a DJ set.
Basically, LALAL.AI is delivering an all-around upgrade with Andromeda. Drums were already one of the top three most separated stems on the platform. LALAL.AI processed over 5 million drum splits in 2025 with a userbase of over 6 million producers, engineers, DJs, and podcasters.
In December 2025, Meta’s independent stem separation benchmark ranked LALAL.AI as the top commercial technology in the Pro category. MusicRadar gave it a 5 out of 5 score for vocal and drum extractions. AAnd in the 2026 Webby Awards, LALAL.AI has won the People’s Vote Webby Award in the Best Use of AI and Machine Learning category
1. Making Remixes and Mashups
If you ever wanted to remix a song by placing the vocals of one track over the music of another song or create a mashup, LALAL.AI is one of the easiest ways to do it. You can upload the song and the AI gives you clean, separated parts.
Thanks to the newly released Andromeda model, you can grab all four parts of a song, like vocals, instrumental, drums, and bass, in one go and get consistent results. You can now remix or create a mashup without any issues.
2. Saving Old or Damaged Recordings
Archival restoration has been one of LALAL.AI’s important features for years and the Andromeda update improves it. The company recently shared the story of a Nashville-based producer who lost recordings in a studio fire. He used LALAL.AI to pull stems out of the only surviving mixed files. So for rescuing lost stems, LALAL.AI is a great platform.
Apart from that, the new update works even better for old songs. Classic recordings from cassettes or old tapes often have muffled or weak bass. The new Andromeda update keeps more of that low-end detail intact. LALAL.AI is one of the best tools if you have any old recordings or favorite tracks from decades ago that you want to clean up.
3. Making Karaoke and Practice Backing Tracks
This is probably the most popular use case for LALAL.AI. People upload a song and remove the vocals to sing karaoke over it. Musicians also use it, but the other way around. For example, you can remove just the drums to practice drumming along with the rest of the song or remove the bass to practice bass.
With the new Andromeda model, these custom practice tracks are faster to make, and they hold up better. Earlier, the drum-removed version of a song would sometimes sound hollow because the AI struggled with louder sections. That’s now improved with the LALAL.AI Andromeda update, and it makes learning an instrument or karaoke singing easy.
4. Learning How Songs Are Made
One of the coolest things you can do with LALAL.AI is to use the tool to learn music. If you’ve ever wondered how your favorite song was recorded and put together, this AI tool lets you hear exactly that. You can hear what the drummer is actually doing, what the bass line sounds like on its own, and how the background vocals were layered.
The new Andromeda model makes this really useful because the isolated parts are now clear enough to study properly. You can do a solo on any part of a song and listen to it on its own, be it the drum pattern, the bass groove, or the guitar riff. It also removes the weird echoes or bleed from other instruments. For music students, this is a fun way to explore music.
5. Cleaning Up Podcasts and Videos
The Andromeda update is mostly about improving the quality of music, but LALAL.AI’s other tools also help podcasters and video creators. The Voice Cleaner tool removes background noise, ambient hum and room noise while the Echo & Reverb Remover fixes recordings that sound hollow and have that bathroom-like echo.
The Andromeda update particularly improves interview or podcast recordings. If you have an interview or podcast that was recorded with music playing in the background, you can pull out the music cleanly so you just get clear voices. Basically, with Voice Cleaner and Echo & Reverb Remover, you have a full audio cleanup tool within LALAL.AI.
LALAL.AI: Pricing and Plans
LALAL.AI has a free Starter tier that gives you 10 minutes to try the service out with no credit card required. After that, there are two paid plans that you can opt for. The Lite plan starts at $7.5 per month (billed annually at $90), and it includes unlimited minutes in the relaxed queue and 90 minutes a month in the faster priority queue.
The Pro plan costs $15 per month (billed annually at $180), and it raises the fast queue to 250 minutes a month and adds the VST plugin for music production software. It also offers local processing, API access, and early access to new features.
With the powerful new Andromeda model, I would strongly recommend users to try LALAL.AI as it’s one of the best AI tools for music production and stem separation.
Shashank Shakya
Partner Content
