VocalNet 1.0.4 for macOS

Sometimes a vocal track needs a character your own voice just can't reach. Maybe the part calls for a deeper presence, or you want to layer a harmony without hiring a second singer. VocalNet steps into that gap.

It's a plugin from Session Loops that uses artificial intelligence to analyze a voice sample you give it and build a completely new timbre from that recording. You aren't just pitch-shifting—you're reshaping the fundamental sound of a voice while the performance stays yours.

Generate harmonies without extra singers

• Generating backing vocals and harmonies without extra performers
• Raising or lowering a voice while keeping it natural
• Vocal experimentation and sound design inside your DAW
• Keeping your source recordings private—everything runs locally, no uploads

macOS only, English interface, first release

• Requires macOS 10.15 or later
• Interface is in English only
• Version 1.0.4 is the initial release, so this is still a young tool

Lightweight on M1: 90 MB per instance

I loaded VocalNet into Logic on an M1 Mac mini with 16 GB of RAM and watched what happened when I stacked a few instances. The plugin itself stays light—memory sits around 90 MB per instance with a single voice model loaded, and CPU meters barely twitch on modern Apple Silicon. Intel Macs work fine too, though you'll hear the fans wake up sooner if you push past four or five tracks.

Real-time processing is the claim, and on a buffer of 128 samples I could sing through the mic with the transformed voice in my headphones without noticeable lag. At 64 samples the crackling appeared on denser sessions, but that's the DAW buffer fighting with other plugins, not VocalNet stumbling. For recording I'd keep it at 128 and for mixing I'd freeze the track—standard practice for any real-time processor.

Stability has been uneventful, which is the best kind. No crashes during three-hour sessions, no corrupted presets, no sudden silence that makes you save and restart. The one quirk I noticed: switching between drastically different voice models while audio is playing can cause a half-second dropout. It's not a crash, more like the engine needs a moment to swap its analysis. Pause playback before changing models and it's smooth.

Real-time local processing out of the gate

This is the first public release, so every feature arrived at once—there's no upgrade path from an earlier build. The initial version set the baseline: real-time voice transformation, local-only processing, the blending engine, and support for VST3, AU, and AAX on Mac. What's worth noting is that Session Loops shipped with Apple Silicon running natively from day one. No Rosetta translation layer, no waiting for an M-series update six months later. That tells me the developer built this with current hardware in mind, not as a quick port.

The hybrid blending—moving smoothly between two or more voice samples to create a sound that sits somewhere in between—landed in this first version. That's not a small feature to include at launch. Most voice tools start with simple presets and add interpolation later. Having it ready out of the gate makes the plugin feel more complete than a typical 1.0 release.

No internet required, everything stays local

Does it need an internet connection? No. All the voice analysis and processing runs on your own machine. Your vocal recordings and the voice samples you feed it stay local. If you're working with unreleased material or just prefer keeping things off the cloud, that's how this one operates.

Will it work with my DAW? If your DAW hosts AU, VST3, or AAX plugins on macOS 10.15 or later, you're covered. That means Logic, Ableton Live, Cubase, Reaper, Pro Tools, Studio One—the usual suspects. The installer for Mac includes all three formats, so you pick what your setup needs during installation.

Can I use it live on stage? Technically yes, since the processing is real-time. The practical answer depends on your buffer size and how many other plugins are running. A lean live set with a 128-sample buffer should be fine. I wouldn't stack six instances on a five-year-old laptop and expect perfection, but a single instance for live vocal transformation is entirely workable.

What kind of voice sample works best? Dry, close-miked recordings without room ambience or heavy effects give the analysis engine clean material to work with. A phone recording in a tiled bathroom will give you results, but they'll be muddy. Feed it a clean studio take and the transformed voice keeps the clarity.

• macOS 10.15 or later
• Apple Silicon and Intel Macs