How do you record Discord audio for a podcast on a Mac?
Craig gives every speaker a separate track, OBS splits you from your guests, BlackHole routes the rest. How each route works on a Mac, what consent looks like, and how to make your own track the strong one.
Three friends record episode one of a tabletop podcast in a Discord voice channel. The host starts QuickTime on her MacBook, talks for ninety minutes, stops, and opens the file. It holds her own voice and nothing else, because QuickTime only listens to a microphone. Episode two she switches to OBS with desktop audio, and this time everyone is there, mixed into one stereo track: when the loudest guest laughs over a punchline, the punchline is gone for good, because no editor can pull two voices apart once they share a track.
Both failures come from the same gap. Discord plays your guests through your speakers and sends your voice out through your mic, and neither path is a recording. To record Discord audio for a podcast you have to put something in the middle, and where you put it decides how much control you have later.
How to record Discord audio: the four routes that work
There are four practical routes on a Mac. They differ in one thing that matters more than price: whether each person ends up on their own track.
| Route | Tracks you get | Cost, as of October 2026 | What your Mac needs | Pick it when |
|---|---|---|---|---|
| Craig bot in the voice channel | One per speaker, in sync | Free (up to 6 hours, files kept 7 days); Patreon tiers $1, $4, $8 a month | Nothing, the bot runs on Craig's servers | Two or more people, you will edit afterwards |
| OBS with macOS Audio Capture | Your mic on one track, all guests mixed on another | Free | OBS 30 or newer, macOS 13 or newer | You also want video, or it is just you and one guest |
| BlackHole plus a recorder, or Audio Hijack | Usually one mixed track for guests, your mic separate if you route it | BlackHole free (donation optional); Audio Hijack $69 | macOS 10.10+ for BlackHole, macOS 14.4+ for Audio Hijack 4.6 | You want system-level routing you control |
| Everyone records themselves | One clean local file per person | Free | Any recorder on each machine | The best possible audio, and guests you can trust with a checklist |
For most podcasts recorded on Discord, Craig is the right default, and the rest of this guide explains why and where it falls short. If your real problem is not saving the conversation but how you sound inside Discord while you talk (thin, crackly, cutting out), start with why your mic sounds bad on Discord instead; that page fixes the live call, this one is about the file.
Before anyone presses record: consent
Craig's own FAQ puts it bluntly: recording someone without their permission is both immoral and illegal. The bot backs that up mechanically. It adds "[RECORDING]" to its nickname in the channel and refuses to record if it lacks permission to change that nickname.
The law depends on where people are sitting. In the United States, federal law (18 U.S.C. 2511) lets a participant record a conversation they are part of. Some states go further: California's Penal Code 632 requires the consent of every party to a confidential conversation. A Discord podcast often spans several states or countries, so the rule that applies is not always obvious, and this article does not try to settle it for your situation.
The practical habit costs ten seconds. Open every session by saying, on the recording, that you are recording for the podcast, and let each guest answer yes. Now the consent is in the file itself, next to the conversation it covers.
Route 1: Craig, one track per person
Craig is a bot you add to your Discord server. It joins a voice channel and records each speaking user to a separate file, and the files stay in sync even when someone joins twenty minutes late. That last part is what makes it a podcast tool rather than a gadget: you can lower the guest who leans into the mic, cut the co-host's cough, and fix crosstalk without touching anyone else.
The full session looks like this:
1. craig.chat > Add to Discord pick your server, approve permissions
2. Join the voice channel yourself
3. /join Craig enters, nickname shows [RECORDING]
4. Record. /note intro done adds a timestamped marker
5. /stop ends the recording
6. Open the link Craig sent you by DM download FLAC, AAC, or an Audacity project
The download link goes by direct message to whoever started the recording, so the person who typed /join is the one who gets the files. Download within 7 days on the free plan. The paid tiers keep recordings for 14 or 30 days, and the $4 tier raises the length limit from 6 to 24 hours.
What Craig actually hears
Craig is a listener in the channel. It records the audio Discord delivers to it, which means each track has already passed through that speaker's Discord settings and their connection. Discord carries voice as Opus at the bitrate set on the channel. Two consequences follow.
First, raise the channel bitrate before you record. Hover the voice channel, click the Edit Channel cog, open Overview, and drag the Bitrate slider up. Server boosts raise the ceiling: 2 boosts give 128 kbps, 7 give 256 kbps, 14 give 384 kbps, according to Discord's Audio Bitrate FAQ.
Second, a guest with a bad connection gets recorded with that bad connection. If their audio drops out on the call, it drops out on their track. Craig has an answer for this: the Craig Webapp. Run /webapp on, start the recording, and share the webapp link from the DM. Guests open it in a browser, and their audio goes to Craig directly, compressed at 128 kbit Opus, bypassing Discord's voice servers and Discord's audio processing. They appear as extra tracks in the same recording.
Route 2: OBS on a Mac
To record Discord audio on a Mac without a bot, use OBS. It is free, and since OBS Studio 30 it no longer needs a virtual cable on macOS 13 or later. The OBS knowledge base describes the macOS Audio Capture source, which can take all desktop audio or a single application.
Set it up so you and your guests land on different tracks:
- Add an Application Audio Capture source and pick Discord.
- Add an Audio Input Capture source and pick your microphone.
- Open Advanced Audio Properties and give the mic Track 1 only, Discord Track 2 only.
- In Settings > Output, switch to Advanced mode and tick both tracks for recording.
- The first time, macOS asks for permission under System Settings > Privacy & Security > Screen & System Audio Recording. Allow OBS, then restart it.
Leave OBS at its default 48 kHz sample rate, the same rate Discord's Opus audio runs at. One quirk reported on the OBS forum: per-application capture can stop after the captured app restarts, and selecting a different app and then Discord again brings it back. If Discord updates itself mid-session, check the meter.
The limit of this route is built in. Discord hands OBS one mix of everyone else, so three guests share Track 2. You get your own voice cleanly separated and nothing more. For the full audio chain inside OBS (levels, filters, monitoring), see the OBS audio settings guide for Mac.
Route 3: BlackHole, or Audio Hijack
BlackHole is a free virtual audio driver. It shows up as an audio device that apps can play into and record from. The classic recipe in Audio MIDI Setup is a Multi-Output Device with Built-in Output on top and BlackHole below it, set as the system output with "Use This Device For Sound Output". You keep hearing your guests, and any recorder that takes BlackHole as its input captures them.
Your own voice is the catch. Discord does not play you back to yourself, so BlackHole alone records only your guests. Getting your mic into the same recording needs an Aggregate Device or a recorder that takes two inputs, and the setup stops being a five-minute job. Whether BlackHole is enough or you need Loopback walks through that routing in detail, including the drift-correction setting that stops glitches in long sessions.
Audio Hijack from Rogue Amoeba is the paid shortcut: $69 as of October 2026, macOS 14.4 or later for version 4.6. Its Application source grabs Discord directly and a separate input grabs your mic, without building devices by hand. If you are weighing Rogue Amoeba's two apps, Audio Hijack or Loopback explains which does what.
On macOS 27 there is one more blunt option: the Screenshot app (Shift-Command-5) has an Include System Audio choice. You get one file with the screen, your mic and everyone else blended. Fine for an archive, poor for an edit.
Route 4: everyone records their own mic
Podcasters call it a double-ender. Discord stays the place where people talk and hear each other, and each person also records their own microphone locally, at full quality, with no network in the path. Afterwards you line the files up on a clap or a counted "three, two, one".
This produces the cleanest tracks of all four routes, and it fails in the most human way: a guest forgets to press record, records the wrong input, or sends a 40-minute file from a 90-minute show. If you go this way, send guests a three-line checklist (input, format, press record before you say hello) and run Craig at the same time as a backup. The Craig tracks are the safety net; the local files are what you publish when everything went right. How to set up Audacity on a Mac covers the local side for anyone on your show who records on a Mac.
Your own track is the one you control
Every route above records your voice as it leaves your microphone. With Craig and the webapp, it passes through Discord or the browser first; with OBS, BlackHole or a local file, it comes straight from the mic input. Either way, nothing downstream adds detail the mic did not capture. A MacBook's built-in mic in a bare room, a headset brushing a collar, a fan behind you: all of it goes into the file, and in an edit it is your track that sits next to a guest's good USB mic.
Discord's own processing helps on the call and has limits on a recording. Its Krisp noise suppression runs on your Mac, and Discord's own Krisp FAQ notes it can reduce quality in a quiet room with a good mic. If you are unsure whether to leave it on, how good Discord noise suppression is covers when it helps.
This is the one part of the chain TunedMic is built for. TunedMic adds a virtual microphone to your Mac, also called TunedMic. You run a one-minute Sound Check (reading a short passage aloud while it takes more than 30 measurements of your voice, your mic and your room), then choose TunedMic as the input in Discord (User Settings > Voice & Video > Input Device), in the OBS Audio Input Capture source, or in your local recorder. Every one of those receives the same processed voice, built from your Sound Check. For podcasts the Creator profiles (Balanced, Broadcast, Crisp) are the place to start. It adds under a tenth of a second of delay and runs locally; your voice is not uploaded anywhere.
The limits are concrete. TunedMic needs a Mac with Apple silicon (M1 or newer) and macOS 14 or later. There is no Windows version, so it improves your track, not the co-host recording on a gaming PC. If you use Bluetooth earbuds, Bluetooth limits how much detail the mic sends, and a wired or USB mic gives it more to work with. And if you already sound the way you want on a dedicated mic in a treated room, you do not need it.
A settings sheet for the session
Keep this next to the keyboard before a recording:
| Setting | Where | Value |
|---|---|---|
| Channel bitrate | Edit Channel > Overview > Bitrate | As high as the server allows (128, 256 or 384 kbps with boosts) |
| Your input device | Discord > Voice & Video > Input Device | Your best mic, or TunedMic |
| Noise suppression | Discord > Voice & Video | One noise reducer in the chain, not two |
| Recorder | Craig /join, guests on /webapp if their connection is shaky | Separate track per person |
| OBS sample rate | Settings > Audio | 48 kHz |
| Download | Craig DM link | FLAC for editing, within 7 days |
| Consent | First 10 seconds of the file | Each guest says yes, on the recording |
When something goes wrong
Craig joins but will not record. Check that it can change its own nickname on your server. Without that permission it stays silent by design.
One guest's track crackles or drops words. Craig recorded what Discord delivered. Ask that guest to join through the Craig webapp next time, or to record locally as a backup.
OBS shows Discord at zero. Check Privacy & Security > Screen & System Audio Recording, then reselect Discord in the Application Audio Capture source.
Your guests hear an echo of themselves. Someone is on speakers. Discord's Echo Cancellation can catch some of it; headphones on every guest fix it at the source.
Your own track sounds worse than the guests'. The recording method is fine; the input is the weak link. That is a microphone and room problem, and the fix happens before Discord, not after.
Do this before your next episode
Add Craig to your server, get one co-host into a voice channel, raise the bitrate, type /join, talk for two minutes, /stop, and download the FLAC tracks. Then listen to your own track alone, with headphones. If it is the weakest of the set and you are on an Apple silicon Mac, watch the 38-second video on the home page: it shows the minute of reading aloud that sets TunedMic up, and Discord is one of the three apps in it. TunedMic is $19 once, and the free 7-day trial starts at first launch with no card or account, so you can run the same two-minute Craig test through it before deciding on pricing.