Key Highlights:
- LALAL.AI launches its first VST plugin, bringing stem separation directly into DAWs like Ableton, FL Studio, and Audacity
- The plugin uses the new Lyra model optimized for local processing on consumer hardware, with six-instrument multi-stem splitting coming soon
- Access requires a LALAL.AI Pro subscription at $15/month, positioning it against free native options in Logic Pro and rising competition from Steinberg
Stem separation exits the browser forever
For five years, LALAL.AI users exported audio, uploaded to a website, waited, downloaded stems, and re-imported files. That workflow dies today.
The company’s first VST plugin eliminates the tab-jumping ritual by placing AI-powered stem separation directly on your channel strip. You now separate vocals from instrumentals without leaving your session.
The plugin works in any DAW supporting VST3 technology. An AU version for Logic users is currently in beta.
Official launch details from LALAL.AI
The Stem Separator plugin currently splits audio into two stems: vocals and instrumental. Multi-stem separation for six instruments is in development.
The takeaway: LALAL.AI built this plugin specifically to keep producers in creative flow rather than managing file transfers.
“LALAL.AI’s VST is not only the best in terms of quality, but it is the only AI-powered VST that truly functions as a VST within a DAW,” said Nik Pogorsky, LALAL.AI’s product owner and co-founder. “Isolating vocals or making a purely instrumental track can happen as fluidly as a producer or artist does any other task inside their preferred software.“
The plugin runs on the Lyra model, engineered for local processing on standard consumer hardware. This differs from LALAL.AI’s cloud-based Andromeda model, which offers higher fidelity but requires server-side processing.
Built-in De-Echo and noise canceling features address common separation artifacts without requiring external repair tools.
The Hardware Shift Making Local AI Possible
The timing of this release connects directly to the NPU revolution in consumer computing.
Intel, AMD, and Qualcomm now ship Neural Processing Units in standard laptop chips. These specialized processors handle music source separation tasks up to 35 times faster than CPUs while leaving your main processor free for other plugins.
This matters for two reasons. First, you get usable latency instead of frozen sessions. Second, your unreleased tracks never leave your machine, solving the data security concerns that kept major labels away from cloud-based separation.
LALAL.AI enters a crowded battlefield. Steinberg SpectraLayers Pro 12 and Logic Pro 11.2 both outperformed LALAL.AI’s Andromeda model in MusicRadar’s recent testing of best AI stem splitters. The new VST uses the lighter Lyra model, suggesting even lower fidelity.
But LALAL.AI’s advantage lies in granularity. Logic Pro splits into four broad categories. LALAL.AI promises specific extraction for electric guitar, acoustic guitar, synthesizer, piano, and wind instruments. For producers sampling a saxophone solo buried in a dense mix, that specificity matters.
The $15/month subscription requirement puts LALAL.AI in direct tension with free native DAW options and the growing list of AI VST plugins competing for your budget.
Your Workflow Changes Starting Now
If you produce sample-based music, this plugin removes the friction between hearing a sound and using it. Browse the full AI tools hub to compare your options.
Consider your DAW plugin integration needs carefully. The Lyra model trades some quality for speed. For archival work requiring maximum fidelity, the cloud-based Andromeda model remains superior.
Mix engineers should update contracts to address stem separation rights. When downstream engineers can un-mix and rebalance your work, the line between mixing and mastering blurs.
Check out Moises AI Studio as an alternative workflow, or explore AI sample finders that complement separation tools.
The stereo file is no longer a finished monument. It is now raw material.”
