What does Voxis actually do?
CEDAR Audio built its reputation restoring decades-old tape and location recordings frame by frame, offline, with all the time in the world. Voxis, released August 27, 2026, does the opposite: it isolates a vocal from a live signal in under 10 milliseconds, fast enough to run on a mix while it's still moving. Instead of a conventional noise gate or a stored noise print, CEDAR Audio runs the signal through a deep neural network trained to recognize human speech and vocals and pull them away from everything else, including background noise, reverb, air-conditioning rumble, traffic and crowd noise. Three linked controls handle low, mid and high frequency bands separately, plus a Voice control to push the isolated signal forward. It ships as VST3, AU and AAX for Windows 10+ and macOS 10.13+, priced at £199 / €229 / $269 plus tax, with a 14-day trial. CEO Andy Trott called it the company's "third new plug-in this year" in its 2026 roadmap, alongside new hardware.
Why does under 10 milliseconds change everything?
Real-time vocal separation has existed for years, RX's dialogue isolate, various live denoisers, but nearly all of them introduce enough latency that you can hear the gap between the source and the processed signal, or they smear transients trying to compensate. Ten milliseconds sits right at the edge of what a monitoring engineer can perceive as delay. That's the number CEDAR is selling Voxis on, and it's aimed squarely at front-of-house consoles, sports and event broadcast, conference AV and worship, places where a vocal has to come out clean while the show is still happening, not after.
What can producers and DJs actually do with it?
Nothing in CEDAR's marketing mentions a DJ booth. But a plugin that strips a vocal off a moving signal in real time doesn't check where that signal came from. Feed it a channel from a mixer with a record playing and Voxis has no way to know it isn't a live microphone: it will try to isolate the vocal the same way. That's the acapella every remixer, edit-maker and mashup producer normally has to beg a label for, or extract offline with something like RipX or LALAL.AI and then fight the artifacts where drums and vocal overlap. Voxis processes live instead of rendering a file, which opens the door to pulling a vocal out mid-set rather than in a DAW the night before. It's a capability, not a workflow: there's no loop point, no stem-export preset, nothing built for a controller.
What's the catch?
Price, for one: £199-plus is a broadcast-engineer's budget line, not an impulse plugin purchase. It's desktop-only, Windows and macOS, no standalone mobile or hardware box, and CEDAR has given zero indication it's chasing the DJ market with this release. Whether that stays true is the actual story here.
