Accentize launches dxSplit, a new dialogue stem separation plugin for post production
Subscribe
X

Subscribe to Mixdown Magazine

24.03.2026

Accentize launches dxSplit, a new dialogue stem separation plugin for post production

dxsplit
Words by Mixdown

German audio software developer Accentize has released dxSplit, a machine-learning-powered plugin that splits dialogue recordings into three separate stems – Voice, Reverb and Noise – giving post production engineers a new level of control over difficult audio.

Cleaning up dialogue has always been one of the more painstaking parts of post-production. Accentize’s new plugin dxSplit takes a different approach to the problem, using neural network algorithms to separate a dialogue recording into three distinct stems so engineers can work with each element independently rather than just applying blanket noise reduction.

Catch up on all the latest news here.

Each stem comes with its own gain control, mute and solo functions, and a dedicated parametric EQ. That means you can shape the voice, pull back room reflections, or dial down background noise without affecting the other elements, all from within the one plugin.

Particularly useful for dialogue editors is dxSplit’s direct integration with dxRevive, Accentize’s speech enhancement plugin. Once the voice has been separated out, dxRevive can be loaded directly within the Voice stem to restore missing frequencies, reduce artefacts and bring back a more natural-sounding result – handy for anything from production dialogue that’s seen better days to archival recordings that need serious attention.

Flexible routing is also part of the package. Each of the three stems can be sent to separate tracks in your DAW via auxiliary outputs, so further processing or mixing downstream is straightforward.

The plugin supports sample rates from 44.1kHz up to 192kHz, runs in mono and stereo, and is available in VST3, AU and AAX formats for both macOS and Windows, with native Apple Silicon support included. Both real-time and offline processing algorithms are on board.

Accentize is based in Darmstadt, Germany, and develops its tools from a background in academic audio processing research. Their existing line-up includes dxRevive, DeRoom, Chameleon and SpectralBalance – all aimed at making complex audio tasks more manageable for working professionals.

dxSplit is available now. A free trial can be downloaded at accentize.com