Free, open-source, node-based audio routing for macOS, Linux and Windows.
Build a node graph of inputs, effects, and outputs. Process any audio source in real time with a flexible processing pipeline.
MIT licensed · no account · no telemetry
Microphone
On macOS and Linux, Splitwave installs a driver and creates virtual audio cables — devices the system treats as real hardware. On Windows, point Splitwave at VB-Cable or a similar virtual cable instead. Any app can play into one, and any app can pick one as its microphone. Both ends land in your graph.
Zoom, Chrome, a DAW
plays into it
Virtual device
real hardware to the OS
Your graph
arrives as a source node
Your graph
processed and mixed
Virtual device
real hardware to the OS
Any app
picks it as its microphone
Virtual devices are available on macOS and Linux.
Twenty-three node types — microphone and system audio capture, noise suppression, EQ, compression, recording — arranged the way audio actually travels. Pick a stage.
Point a Plugin node at anything already installed on the machine — CLAP, AU or VST3 — and wire it into the chain like a built-in effect. Same cables, same signal.
Plugin node
hosts it in the graph
The rest of your chain
meters, outputs, peers
A plugin is first offered the node's full channel count, and one instance handles everything. If it refuses that layout, the signal runs as stereo pairs with an instance per pair, and every instance follows the same parameter changes. Audio Units are a macOS format, so AU hosting is macOS-only.
A cable can run to another computer. Whatever comes back shows up as an ordinary source node, no different from a microphone.
THIS MACHINE
Your graph
inputs, effects, outputs
UDP · WebRTC
ANYWHERE ELSE
Peer receiving your mix
Peer sending theirs back
Splitwave is a desktop audio router. Instead of a fixed mixer layout you build a node graph: drag in the sources you want (microphone, system audio, a single app, a file), wire them through effect nodes, and end at outputs such as speakers, a recording file, or a virtual device other apps can read. The graph runs in real time and you can rewire it while audio is playing.
Open the sidebar, drag an input node and an output node onto the canvas, and connect the output handle of one to the input handle of the other. Add effect nodes in between by wiring them into the chain. Press Activate to run the graph, and Stop to tear it down again.
A virtual device is a fake sound card that other applications can select as an input or an output. Send Splitwave audio into a virtual output, then pick that same device as the microphone in Zoom, Discord, OBS, or your DAW — those apps receive your processed signal as if it came from real hardware. On macOS and Linux, Splitwave creates as many virtual devices as you need and lets you name them yourself. On Windows it does not create them — install VB-Cable or a similar virtual audio cable and Splitwave will use it like any other device.
Yes, and that is the common case. Build a chain such as microphone into noise suppression, gate, compressor, and EQ, end it at a virtual output device, then select that device as the microphone inside the app you are calling or streaming from. You can also branch the same processed signal into a File Recording node to keep a local copy.
Yes. Any output handle can be wired to several nodes, so one microphone can feed your headphones, a recording file, and a virtual device simultaneously, each through its own chain of effects. Several sources can also be merged into one output.
Splitwave runs on macOS 13 Ventura and newer (Apple Silicon and Intel), Linux (x86_64, with a PipeWire audio session), and Windows 10 and 11 (64-bit). On macOS, system audio capture uses ScreenCaptureKit, which Apple introduced in macOS 13.
No kernel extension and no System Integrity Protection workaround. Splitwave does install a small userspace audio driver to expose virtual devices — you will be asked for your macOS password once during install, and again every time you add or remove a virtual device, because macOS requires administrator approval for any change to the system audio configuration.
Only the standard permissions for the sources you actually use. On macOS that means microphone access for mic capture, screen recording for system or per-app audio capture, and disk access when you choose a recording destination. Windows and Linux ask for microphone access in the same way, through their own system prompts.
Yes. Add a File Recording node anywhere in your pipeline and Splitwave will write a lossless WAV, FLAC, or AIFF file, or an encoded Opus, MP3, or AAC stream, while the rest of the graph keeps playing back to your speakers, headphones, or virtual outputs.
Yes. Create a virtual output device in Splitwave (or use VB-Cable on Windows) and pick it as the input device in Logic Pro, Ableton Live, Reaper, GarageBand, or any other DAW. You can also route your DAW master back through Splitwave for live effects.
Splitwave resamples between them, so a 44.1 kHz source can feed a 48 kHz output without you configuring anything. The same applies inside the graph: the DeepFilterNet noise suppressor runs its model at 48 kHz, and resamples in and out automatically when your pipeline is running at some other rate.
Yes. Splitwave hosts third-party CLAP, AU, and VST3 plugins as effect nodes: point a Plugin node at an installed plugin and wire it into the chain like any built-in effect. The plugin's editor window opens from the node, and you can also expose and automate its parameters directly on the node.
Yes, at two levels. Named pipelines keep whole graphs — every node, connection, and setting — so you can switch between a podcast setup and a gaming setup. Separately, effect nodes have their own presets: the compressor, gate, limiter, EQ, reverb, delay, saturator, and noise suppressor each store named parameter sets, and Splitwave ships with built-in ones you can select but not overwrite.
No. All processing happens locally on your machine, and there is no account, no analytics, and no cloud rendering. Audio leaves your computer only when you deliberately add a network node — WebRTC or the UDP sender — and point it at a peer yourself. The app does contact GitHub to check for new releases, which is the only routine outbound request it makes.
None of the three systems ships an easy way to send one app's sound into another app. Splitwave fills that gap the same way everywhere: capture a single application with the App Audio node, wire it through any effects you want, and end the chain at a virtual output device. The receiving app — Discord, OBS, QuickTime, your DAW — just selects that device as its input, so the audio never leaves your machine and no extra hardware is involved. Splitwave creates those virtual devices itself on macOS and Linux; on Windows, install VB-Cable or a similar virtual cable first.
It covers the same ground and then some. Tools like BlackHole and VB-Cable give you a plain virtual audio cable with no processing; Loopback and VoiceMeeter add a mixer on top. Splitwave gives you a full node graph on macOS, Linux and Windows — noise suppression, EQ, compression, third-party VST3/AU/CLAP plugins, recording, and network streaming — free and MIT licensed, plus unlimited named virtual devices of its own on macOS and Linux. On Windows it works alongside VB-Cable rather than replacing it.
Add a System Audio node — it captures everything the machine is playing, using ScreenCaptureKit on macOS 13+, and the system's own loopback capture on Windows and Linux — and wire it into a File Recording node. You can record to WAV, FLAC, AIFF, Opus, MP3, or AAC, mix your microphone into the same file, and keep monitoring through your headphones while the recording runs.
Splitwave is free and open source under the MIT license. Source code lives on GitHub — you can build it yourself, audit the audio path, or contribute back.