Podcast recording with multiple mics, every voice on its own track

The goal: one track per voice, forever editable

When three people share one recorded track, every edit is a compromise. You cannot turn down the host who leans in without turning down the guest who whispers, you cannot cut one person's chair squeak without cutting the punchline underneath it, and you cannot fix a too-quiet voice in post without dragging everyone else up with it. When each voice sits on its own track, all of those become one-lane fixes. It is the same principle streamers use to keep game audio and voice on separate tracks: separate at the source, and every later decision stays open.

Getting there takes four layers that have to agree with each other: each mic needs its own physical input, each input needs its own channel strip, each strip needs its own bus, and your recording app needs to capture each bus as a separate track. One clean wire per voice, end to end. Here is how to build it on Windows with patchd.

Step 1: give every mic its own physical input

Separation starts at the hardware. If two mics arrive on one input, no software can ever pull them apart again, so count inputs first. For XLR mics, that means an audio interface with at least as many mic inputs as you have people: a two-input interface for two hosts, a four-input interface for a round table. The interface handles preamps and conversion; patchd runs over it (ASIO first) and adds the routing and processing, so the two are layers, not rivals. If you are weighing that split, the interface versus software mixer guide covers it.

USB mics work too, because each one shows up in Windows as its own device. Two USB mics make a perfectly good two-person setup. Past two, a multi-input interface is usually the saner buy: every mic shares one clock and one driver instead of several USB devices each doing their own thing.

Step 2: one strip per voice in patchd

In patchd, every input you add becomes a strip with its own fader, live meter, mute, and effects rack. Add each mic as its own selectable strip, so each voice gets its own level and its own processing. Honest tier math before you plan a panel show: the free tier gives you two selectable inputs total, which covers a two-mic podcast with no cable to spare. The Studio tier raises that to five selectable inputs in total, enough for four or five people each on their own strip.

Two strips are always there regardless: Desktop for system audio and Communications for chat apps. That second one matters for podcasts: a remote guest calling in through a chat app arrives on the Communications strip, which means the remote voice can travel its own lane exactly like the mics in the room.

Step 3: clean up each voice on its own strip

Per-strip processing is the quiet payoff of this layout. Three people never need the same settings: one sits close to a dynamic mic in a treated corner, another is two feet off a condenser next to a window. Open each strip's rack, add nodes from the + Add Node menu, and tune per voice. A noise gate set against that person's room, so their track goes silent between their lines, and an EQ shaped to that person's voice, a high-pass around 80 to 120 Hz for rumble and small moves from there. The walkthroughs in setting up a noise gate and EQing your mic apply per strip, per person. Because the processing lives on the strip, it is baked into that voice's track and nobody else's.

Step 4: route each strip to its own bus

A bus is a virtual output that other apps see as a microphone, and this is where the lanes get drawn. Each strip has a row of colored routing pills under its BUS header. Light exactly one bus per voice: host on Bus 1 (cyan), co-host on Bus 2 (violet), guest on Bus 3 (magenta). A glance at the colors confirms no two voices share a lane. The free tier includes three buses, which pairs with its two mic strips and leaves one spare; Studio raises that to six buses in total, so five strips can each have their own lane with room left over.

Note what you are not doing: patchd strips have one fader each, and destinations differ by which sources you route there, not by separate per-destination levels. That constraint is a feature here. Set each voice's level once on its strip, and every bus carries exactly one voice at that level.

Step 5: record the buses as separate tracks

patchd routes; your recording app records. Each patchd bus presents itself to Windows as a microphone, so open your recording software, a DAW or a multitrack recorder, create one track per voice, and set each track's input to the matching bus: track one records Bus 1, track two records Bus 2, track three records Bus 3. Hit record and you are capturing clean, separately processed voices, one per track, ready for the edit. Any app that can record from a microphone can record a patchd bus, because to the app that is all a bus is.

Step 6: let everyone hear the show

Monitoring rides the hardware outputs, which are separate from the buses. Route all the mic strips to your headphone output so the people in the room hear each other, and split that output to everyone's headphones with your interface's headphone outs or a small headphone amp. Latency through patchd is about 10.7 ms through the bus at a 512 buffer, and smaller ASIO buffers take it lower; the buffer itself is set in your interface's own control panel, and patchd's master clock readout shows the live result. If self-monitoring feels off, see monitoring your mic with low latency.

Where patchd is today

patchd is built for exactly this shape of problem: one strip per source, per-strip processing, color-coded buses your recorder captures as separate tracks. It is in development now. Join the waitlist and we will tell you the moment it is ready to install.