LALAL.AI is moving beyond browser-based audio processing with its first embedded hardware integration, bringing real-time stem separation directly to an offline device.
Unveiled at IBC2026 in Amsterdam, the company’s stem-separation engine now runs natively inside Neko, a compact wireless AI studio developed for guitar players by Neko Engineering.
The integration can separate a track into vocals, drums, bass and guitar in real time without an internet connection. Processing takes place locally, removing cloud latency, recurring inference costs and the need to upload audio to an external server.
For six years our tech has lived in a browser tab. Putting it inside a physical device, running in real time with nothing sent to a server, is a different kind of product decision. It’s a signal that separation-quality AI doesn’t have to depend on the cloud to be useful. – LALAL.AI co-founder Nik Pogorsky
Neko Engineering CEO Vadim Smirnov said the priority was creating stem separation musicians could use during live playing without waiting for a network connection.
Alongside the hardware debut, LALAL.AI is demonstrating applications designed for broadcast, post-production, dubbing and localization workflows.
These include extracting dialogue when a separate music-and-effects track is unavailable, cleaning location recordings affected by wind or traffic, reducing room reverberation and recovering stems from finished mixes without access to the original multitracks.
The company is also showing how its Voice Cleaner can remove background music while retaining commentary and ambient sound. This could help production teams prepare social and highlight footage affected by automated Content ID systems, although it does not replace the need for proper music rights and clearance.
For studios working under strict security or confidentiality requirements, LALAL.AI offers local processing through its Lyra-powered VST3 and AU plugin. The software can operate inside platforms including Ableton Live, Reaper, DaVinci Resolve Fairlight and Adobe Premiere Pro without source material leaving the user’s system.
The Neko integration and the demos on our stand are really the same story told twice. The problem is the same: getting usable, controllable audio out of something that wasn’t handed to you clean. – LALAL.AI Business Development Manager Anastasia Shikina
The LALAL.AI-powered Neko device is available for pre-order at $349, ahead of a planned Kickstarter launch in October 2026 and shipping scheduled to begin in 2027.
For audio AI, the shift towards edge processing is significant. Running separation locally can improve speed, privacy and reliability while opening the technology to live performance and embedded hardware.
The next test will be whether these models can maintain professional-quality results consistently outside the cloud.