How to Mix a Multi-Speaker Podcast in REAPER (Stock Plug-Ins Only)

Four people record in four different rooms, on four different microphones, and you get four files with four different noise floors. Here is the full REAPER routing and processing chain that turns that into a broadcast-ready episode — using nothing but plug-ins that already ship with REAPER.

Why one track per speaker is not optional

The most common shortcut is to drop every recording onto one track and hit it with one chain. It sounds acceptable in a quiet room and falls apart everywhere else. Reasons:

• Noise floors differ. One guest has a laptop fan, another has a fridge, a third has street noise. A single noise profile cannot remove three different noises — it can only target one, and it damages the other voices trying.
• Levels differ. Someone sits 10 cm from the mic, someone sits 40 cm away. In one shared chain the quiet guest gets compressed against a threshold set for the loud host.
• Tone differs. A dynamic mic in a treated room and a USB mic in a kitchen do not want the same EQ or the same compression.

So the session is built around one voice track per person, each with its own independent chain, all feeding a shared dialogue bus.

Step 1 — Build the session skeleton

Four buses, in this order from top to bottom:

MASTER
MUSIC BUS
  MUSIC          (empty — your background music)
DIALOGUE BUS
  HOST
  GUEST 1
  GUEST 2
  GUEST 3
SFX BUS
  SFX            (empty — your transitions)

In REAPER, MUSIC BUS, DIALOGUE BUS and SFX BUS are folder tracks: set the bus track as a folder parent (the small folder button in the lower-left corner of its control panel), then indent the child tracks beneath it. Folder tracks give you three things for free: a single fader per group, a single place to insert group processing, and — critically for the next step — a single point to send from.

Put MASTER at the top only if that is how you like to see it; REAPER does not care about visual order, and the master track is always the hardware output.

Step 2 — Align the takes before you process anything

Remote recordings almost never share a time origin. Even when a platform exports “synchronised” files, each client starts recording at its own moment, so the files will be offset by anywhere from a few hundred milliseconds to several seconds.

Do not trust the file start. Find a moment where everybody speaks at once — the “hey, can you hear me?” at the top is usually perfect — and slide the guest clips until the waveforms line up at that point. Zoom in on the transient and match the attack, not the tail.

Step 3 — Clean up: noise reduction with ReaFir

REAPER's ReaFir in Subtract mode is a noise-profile-based reducer: it learns the spectrum of a steady noise, then subtracts that spectrum continuously. This is the same basic principle as iZotope RX's spectral denoise, minus the manual editing tools.

1. Put ReaFir as the first plug-in on each voice track, set the mode to Subtract.
2. Find 10–20 seconds where the speaker is not talking — room tone only. Every decent recording has it at the head.
3. Play that section, tick Automatically build noise profile, let it run for a couple of seconds, then untick it.
4. Repeat per speaker, per episode. The profile is a fingerprint of that room and that microphone on that day; a profile from last week's session will not fit.

Step 4 — Level the voices with ReaComp

This is the step that makes the biggest audible difference, and the one most home mixes skip. Each speaker track gets a ReaComp configured as a slow, gentle auto volume rider:

• Detector input: main input, RMS rather than peak — you want to follow perceived loudness, not transients.
• Knee: soft (a wide knee). Hard-knee compression on speech is audible as pumping.
• Release: slow, in the hundreds of milliseconds. Fast release makes the gain chase every syllable and creates a breathing artefact.
• Auto make-up gain: off. Leave it off deliberately.

The auto make-up point matters. Its job is to restore the output to the input level — which cancels out exactly what you just did. With it off, the output gain becomes your inter-speaker level control: trim each track's output so the host and every guest sit at the same perceived loudness, and the difference is baked into the session. That is a one-time setup per template, not a per-episode task.

Step 5 — Duck the music automatically (ReaComp sidechain)

Background music should drop when someone speaks and come back when they stop. Doing that by hand with volume automation is 40 minutes of clicking per episode. REAPER can do it with a second ReaComp and one send.

1. DIALOGUE BUS  →  I/O  →  Add send to MUSIC BUS
     Audio  1/2 → 3/4     (post-fader)

2. MUSIC BUS     →  insert ReaComp
     Detector input:  Auxiliary Input L+R

Now the dialogue bus feeds the compressor's detector but not its audio path: when the voices get loud, the music is turned down; when they stop, it recovers. Set the ratio high and the release moderate so the music breathes back rather than snapping.

Auxiliary channels 3/4 are used rather than 1/2 so the music bus is not receiving the dialogue as audio — the voices must not be heard twice.

Step 6 — Master limiting with ReaLimit

Put ReaLimit on the master. Two controls matter:

• Ceiling: −1.0 dBTP. This is the delivery standard for every major podcast platform and streaming service.
• Threshold: close to the ceiling. The limiter is a fuse, not a compressor. If the threshold sits several dB below the ceiling it will flatten normal peaks and audibly squash the mix. Set it just under the ceiling so it only catches the occasional transient.

REAPER's limiter has no output gain, by design: the ceiling is the output level. Do not go looking for a gain knob — overall loudness is handled at render, in the next step.

Step 7 — Render to −16 LUFS

Open File → Render and configure it once, then save the project — REAPER stores the render settings inside the .RPP file, so they travel with the session instead of living in your local preferences.

Output directory: leave it empty. If you hard-code a path it will be written into the project file and break on anyone else's machine — including yours, if you move the session. REAPER asks for a folder on first render and remembers it afterwards.

Because the loudness normalisation happens at render, the per-track compressor output gain only has to make the speakers match each other. The whole programme level is set once, at the end, to the platform standard.

What stock plug-ins cannot fix

Being honest about the limits saves time. None of these are recoverable in REAPER's built-in set:

Recording checklist (fix it at the source)

• One file per person, original WAV — never an audio file sent through WhatsApp, WeChat or any messenger (those re-encode it).
• At least 3 dB of headroom; never let the input meter touch 0 dB.
• 20 seconds of silence at the top of every recording — that is your noise profile.
• Mic roughly a fist away, slightly off-axis, pop filter on.
• Everyone records locally, not through the call platform's recorder.

Build it yourself, or start from a finished template

Everything above is enough to build this session from scratch. Here is what that costs, honestly:

If you enjoy building sessions, build it — the guide above is complete. If you would rather spend those hours recording, the template is the same session, already finished.