Audio setup and latency
Device types, buffer sizes and input gain, the settings that decide whether a rig feels like an amp or like a video call.
Latency is the whole game for a live rig. Everything on this page exists to get the round trip, string to speaker, low enough that you stop noticing it, and stable enough that it never crackles mid-song.
Pick the right device type
On Windows, the device type matters more than any other setting:
| Type | Use it? | Why |
|---|---|---|
| ASIO | Yes, always, if your interface has a driver | The only Windows path built for low round-trip latency. |
| Windows Audio (WASAPI, exclusive) | Workable fallback | Higher latency than ASIO, but usable if no ASIO driver exists. |
| DirectSound / MME | No | Latency measured in tens of milliseconds. Fine for playback, unplayable for a rig. |
Install your interface’s own ASIO driver from the manufacturer. A generic one wrapping a device that has no native support will not beat WASAPI.
On macOS there is nothing to decide. CoreAudio is the native low-latency driver and your interface talks to it directly.
On Linux, choose JACK. On a current desktop that is PipeWire’s JACK support (the pipewire-jack package), which lets Plectrify share the interface with everything else while
keeping low latency, and Plectrify finds it by itself: no pw-jack wrapper and no loader
configuration. ALSA opens the interface exclusively, so every other application on the
desktop goes silent for as long as Plectrify runs; it is the fallback for a machine with no
PipeWire and no JACK server. The section Linux: PipeWire and JACK below has the two system settings worth checking.
Buffer size
The buffer size is the trade you actually control: smaller buffer, lower latency, more risk of dropouts. Your CPU has to render every plugin in the chain inside one buffer’s worth of time, every time, without ever being late.
A reasonable way to land on a number:
- Start at 128 samples.
- Play hard for a minute with your full rig loaded, not an empty one. A heavy amp sim and a convolution reverb cost far more than a tuner.
- If you hear clicks, crackles or dropouts, go up one step (256, then 512).
- If it is clean and you want tighter feel, go down one step and repeat.
At 48 kHz, 128 samples is roughly 2.7 ms of buffering each way. 256 is about 5.3 ms. Most players stop being able to feel the difference somewhere around 10 ms round trip.
The status bar shows the engine’s live load. If it is regularly near the top, the fix is a bigger buffer or a lighter chain, not a smaller buffer.
Linux: PipeWire and JACK
On a PipeWire desktop, Plectrify is one JACK client among the graph’s other clients, and the interface is driven by PipeWire itself. The buffer size you pick in Plectrify is what PipeWire runs the graph at; the sample rate is PipeWire’s and is shown rather than offered. Two things outside the app decide whether that graph runs cleanly, and both are worth checking once. The numbers below were measured with a Focusrite Scarlett Solo (3rd gen) and a five-module rig, which ran clean at 64 samples after both were in place.
Which side of a USB interface drives the clock. PipeWire can drive a duplex interface from its capture side or its playback side, and by default prefers capture. On a USB interface that takes its clock from the host (Focusrite’s 3rd-gen range among them) the playback side then runs as a follower whose buffer keeps running dry, and every time it does PipeWire resyncs it, which you hear as a click. The symptom is unmistakable in PipeWire’s own log:
journalctl --user -u pipewire -f
spa.alsa: hw:USBp: follower avail:1 delay:1 target:384 thr:256, resync (11 suppressed) The p after the card name says it is the playback side. The fix is one WirePlumber rule
that makes playback the driver. Find your interface’s sink name with wpctl status (it starts
with alsa_output.usb-), then create ~/.config/wireplumber/wireplumber.conf.d/51-interface-driver.conf:
monitor.alsa.rules = [
{
matches = [ { node.name = "~alsa_output.usb-Focusrite_Scarlett.*" } ]
actions = { update-props = { priority.driver = 2200 } }
}
] Adjust the match to your interface, then systemctl --user restart wireplumber and reopen the
device in Plectrify. The resync lines stop.
PipeWire’s own real-time priority. PipeWire asks for real-time scheduling when it starts, and on some systems it starts before the service that grants it, so its processing thread runs at ordinary priority for the whole session. Plectrify’s own audio thread is unaffected, but the thread that wakes it is late, and the block budget shrinks by however late it was. Check with:
ps -L -o cls,comm -p $(pgrep -x pipewire) The data-loop.0 thread should show RR. If it shows TS, add yourself to the pipewire group so the system’s limits let it set the priority itself at boot (sudo usermod -aG pipewire $USER, then log out and in), or for the current session systemctl --user restart pipewire pipewire-pulse wireplumber and reopen the device.
Watching it. pw-top shows every node in the graph once a second; Plectrify is the JUCEJack row. Its ERR column counts the cycles where the app’s block was not ready in time,
and it should stay put while you play. If it climbs, the buffer is too small for the chain on
this machine, or one of the two settings above is still off.
Input gain and the meters
Set input gain at the interface first, and use Plectrify’s input gain only for trim.
Aim for peaks well below clipping. Amp sims are built expecting a signal at roughly the level a real pickup sends into a real amp, and a hot input does not make them sound bigger. It makes them sound wrong, and it clips before the plugin ever gets a say. The meter turns amber as you approach the ceiling and red when you hit it; red means you have already lost the transient.
The mono fan-out
A guitar is mono, so the first enabled input channel is copied to both sides of the rack. This happens before the first slot, which is why a stereo reverb at the end of your chain still gives you a proper stereo image.
If you have two instruments plugged into one interface, the channel you enable first is the one that gets played.
Latency you cannot fix in software
Two costs sit outside the app: your interface’s own converters, and any wireless system between guitar and interface. Neither shows up in a buffer-size setting, and both are real. If the round trip feels wrong even at a small buffer with a light chain, the interface is usually the constraint.