The first attempt at audio taught one lesson: a show’s audio is not a row of faders. It is a set of signals that travel along different paths, get processed differently and end up in different places, often at the same time.
So the second attempt started from that idea.
From mixer strips to a graph
In the familiar model, each source has a fader and a few send buttons, and a fixed set of buses sends sound out. It is easy to understand, and it breaks as soon as a show gets interesting.
In Muxivo Studio, audio is a routing graph. Sources feed buses. Buses can feed other buses or named outputs. The same source can reach several places with different levels. The graph is checked as you edit it, so you cannot build a loop that feeds back into itself.
And it follows the picture. Audio knows which scene is on air: a scene’s sources can follow the video in and out, while a microphone set to stay on stays on through every cut. Preview is isolated, so a video waiting in Preview makes no sound on Program until it is taken live.
Processing on the right path
Where you put an effect matters as much as which effect it is.
Muxivo Studio lets you attach a chain of processing to four kinds of path, each with a clear owner:
- A layer, when one use of a source needs treatment, say a guest’s microphone in one scene only.
- A source, when every use of it should be treated the same.
- A bus, when a whole group needs shaping together, with a group processor such as a compressor on the bus.
- A sidechain, for processing driven by another signal.
Chains have built-in effects, ordering, bypass and reset, and changes are applied safely while audio is running, without clicks or dropouts.
Ducking that does not get in the way
A classic live need: music should dip when someone talks, or when an overlay with its own sound comes on air.
Here, ducking is an independent sidechain stack. It lowers the music without replacing or disturbing the processing you already have on that bus. Your compressor keeps doing its job; ducking is added on top and removed cleanly.
Automix joins it for panels with several microphones: it balances the voices automatically so the operator does not have to ride every fader.
Plugins that cannot take the show down
Professional audio lives on plugins, so Muxivo Studio hosts VST3 effects with their real state, parameters and editors.
The important word is hosts. Each plugin runs in its own supervised process, separate from the application. If a plugin crashes, it crashes alone. The show keeps running, and the failure is reported instead of freezing everything.
Soundcheck
This is the part I am most pleased with.
Before something goes on air, you want to hear it. Is the guest’s microphone clean? Does the noise reduction on that source actually help? Is the video’s audio at the right level?
Soundcheck lets you listen to a source, a layer or a bus privately, on a separate output such as your headphones, while Program carries on untouched. Plugins on the path you are checking keep their own independent state during Soundcheck, so adjusting them privately never disturbs what the audience hears.
It is a small idea with a big effect on nerves during a live show.
Getting it out
Audio leaves the same way the picture does: to devices, and over the network. Buses can be sent out as NDI audio, so other machines and software receive a clean mix. Levels are metered along the way, and audio paths can be recorded.
What is next for audio
The graph was designed for what comes next.
- Mix-minus: a remote guest should hear everyone except themselves. In a graph this is a property of a return path, not a workaround, and the goal is for every guest to get their own automatically.
- Talkback: a director speaking privately to a host or guest, through the same graph.
- More inputs, such as audio from a single application or the whole desktop.
Most of what this chapter describes runs today and is being hardened through real use. The guest-related parts are still on the plan.
Picture and audio now had solid foundations. The third kind of signal in a modern show is data: chat, polls, scores and alerts. That is the next chapter.