audiotee/README.md

AudioTee

⚠️ API Instability Warning: The AudioTee API is unstable at present and subject to change without notice.

AudioTee captures your Mac's system audio output and writes it in PCM encoded chunks to stdout at regular intervals. All logging and metadata information is written to stderr, meaning at its simplest you can capture whatever's playing through your speakers to a file like this:

/path/to/audiotee > output.pcm

It's more likely you want to capture this output programmatically. Check out AudioTee.js for a simple Node.js package which does this.

System audio is captured using the Core Audio taps API introduced in macOS 14.2 (released in December 2023). You can do whatever you want with this audio - save it to disk, visualise it, transcribe it, etc.

By default, AudioTee captures audio output from all running processes. Tap output defaults to mono (configurable via the --stereo flag) and preserves your output device's sample rate (configurable via the --sample-rate flag). Only the default output device is currently supported.

My original (and so far only) use case is streaming audio to a parent process which communicates with a realtime ASR service, so AudioTee makes some design decisions you might not agree with. Open an issue or a PR and we can talk about them. I'm also no Swift developer, so contributions improving codebase idioms and general hygiene are welcome. I have internal variations (and, frankly, improvements) of AudioTee which allow recording mic input as well as system audio, and I'm open to making that part of the main API.

Why?

Recording system audio is harder than it should be on macOS, and folks often wrestle with outdated advice and poorly documented APIs. It's a boring problem which stands in the way of lots of fun applications. There's more code here than you need to solve this problem yourself: the main classes of interest are probably Core/AudioTapManager and Core/AudioRecorder. Everything's wired together in CLI/AudioTee. The rest is just CLI configuration support, output formatting logic, and some utility functions you could probably live without.

Requirements

• macOS 14.2 or later
• Swift 5.9 or later (no need for XCode)
• System audio recording permissions (see below)

Quick start

The following will start capturing audio output from all running programs and write binary chunks of raw PCM audio data to your terminal:

git clone git@github.com:makeusabrew/audiotee.git
cd audiotee
swift run

More usefully, you can redirect stdout to a file:

swift run audiotee --sample-rate 16000 > output.pcm

Which you can play back using something like ffplay:

ffplay -f s16le -ar 16000 output.pcm

Build

# omit '-c release' to get a debug build
swift build -c release

Usage

Basic usage

Replace the path below with .build/<arch>/<target>/audiotee, e.g. build/arm64-apple-macosx/release/audiotee for a release build on Apple Silicon.

# Write raw PCM audio to stdout (logs go to stderr)
./audiotee

# Redirect audio to a file
./audiotee > output.pcm

# Pipe to another program
./audiotee | your_audio_processing_tool

# Redirect logs as well
./audiotee > captured_audio.pcm 2> audiotee.log

Audio conversion

Note that performing any sample rate conversion will also convert the output bit depth to 16-bit - assuming an original depth of 32-bit this results in a loss of dynamic range in exchange for a 50% reduction in output size. For ASR services, 16-bit is sufficient, but it's a non-obvious behaviour worth being aware of.

# No sample rate preserves your device's default (probably 44.1 or 48kHz with 32-bit float bit depth)
./audiotee

# Any sample rate (even one matching your device default) converts to 16-bit signed integers (half the bandwidth)
./audiotee --sample-rate 16000

# Other supported sample rates: 22050, 24000, 32000, 44100, 48000
./audiotee --sample-rate 44100

Tap configuration

For now, only a subset of the CATapDescription (https://developer.apple.com/documentation/coreaudio/capturing-system-audio-with-core-audio-taps) interface is exposed. PRs welcome.

Note that trying to include or exclude a PID which isn't currently playing audio will probably fail to convert to an Audio Object and will cause the process to exit.

# Tap all system audio (default)
./audiotee

# Tap only a specific process (by PID)
./audiotee --include-processes 1234

# Tap multiple specific processes
./audiotee --include-processes 1234 5678 9012

# Tap everything *except* a specific process (by PID)
./audiotee --exclude-processes 1234

# Exclude multiple specific processes
./audiotee --exclude-processes 1234 5678 9012

# Mute processes being tapped (so they don't play through speakers)
./audiotee --mute

# Custom chunk duration (default 0.2 seconds, max 5.0)
./audiotee --chunk-duration 0.1

Microphone capture

AudioTee can optionally capture the default input device (microphone) as a second, independent track alongside system audio — useful for "me vs them" diarization. Mic audio is written to its own file rather than stdout, since interleaving two live PCM streams from separate Core Audio IO threads onto one stream would corrupt both.

# Capture system audio to stdout and mic audio to a separate file
./audiotee --capture-mic --mic-output mic.pcm > system.pcm

--sample-rate and --chunk-duration apply to both tracks. On stderr, each track's metadata message carries capture_mode: "audio" or "mic", and its stream_start/ stream_stop messages carry "audio"/"mic" as their data value — so you can tell which track a given message belongs to when both are interleaved in the same log.

Output

AudioTee writes raw PCM audio data directly to stdout in chunks. All logging, metadata, and status information is written to stderr.

Audio format

• Format: Raw PCM audio data
• Channels: 1 in Mono mode (default), 2 in stereo mode
• Sample rate: Matches your output device's sample rate by default (configurable)
• Bit depth: 32-bit float by default, or 16-bit when sample rate conversion is performed
• Endianness: Little-endian
• Chunk duration: 200ms by default (configurable)

Logs and monitoring

All program logs are written to stderr and can be captured separately:

# Capture audio and logs separately
./audiotee > audio.pcm 2> audiotee.log

# View logs in real-time while capturing audio
./audiotee > audio.pcm 2>&1 | grep "AudioTee"

Command Line options

• --include-processes: Process IDs to tap (space-separated, empty = all processes)
• --exclude-processes: Process IDs to exclude (space-separated, empty = none)
• --mute: Mute processes being tapped
• --stereo: Record in stereo
• --sample-rate: Target sample rate (8000, 16000, 22050, 24000, 32000, 44100, 48000)
• --chunk-duration: Audio chunk duration in seconds [default: 0.2, max: 5.0]
• --capture-mic: Also capture the default input device (microphone) as a second track
• --mic-output: File path to write microphone PCM audio to (required with --capture-mic)

Permissions

There is no provision in the code to pre-emptively check for the required NSAudioCaptureUsageDescription permission, so you'll be prompted the first time AudioTee tries to record anything. Note that some terminal emulators like iTerm don't always prompt for these permissions (though the macOS builtin terminal definitely does), so you might need to grant them ahead of time if audiotee runs but never records anything.

--capture-mic requires the standard, separate Microphone TCC permission (not NSAudioCaptureUsageDescription), and will trigger its own first-run prompt.

If you want to check and/or request permissions ahead of time, check out AudioCap's fantastic TCC probing approach.

Stable permissions across rebuilds

By default, swift build ad-hoc-signs the binary, and ad-hoc signatures are keyed off the binary's own hash — so every rebuild looks like a new, untrusted app to TCC and you get re-prompted (or worse, silently record silence). swift run also invokes the binary from inside .build/, and TCC has been observed keying on binary path too, which causes the same problem across rebuilds even without touching signing.

scripts/build-signed.sh builds a release binary, signs it with a stable identity (a free self-signed certificate in your Keychain — no paid Developer ID needed for personal use), and installs it to a fixed path (~/bin/audiotee by default). The binary also embeds an Info.plist at link time (see Package.swift) carrying a fixed CFBundleIdentifier plus NSAudioCaptureUsageDescription/NSMicrophoneUsageDescription, without needing a full .app bundle — this matters if you invoke audiotee as a subprocess from another program (e.g. a Python ASR orchestrator) rather than through Launch Services.

One-time setup — either via the GUI (Keychain Access → Certificate Assistant > Create a Certificate..., Identity Type "Self Signed Root", Certificate Type "Code Signing", then set that certificate's Trust > Code Signing to "Always Trust"), or entirely via CLI with scripts/create-signing-identity.sh (review it first — it generates a key, imports it into your login keychain, and trusts it for the codeSign policy). Either way, after that:

scripts/build-signed.sh              # build, sign, install to ~/bin/audiotee
scripts/build-signed.sh --reset-tcc  # also reset TCC state — useful after changing
                                      # Info.plist or the signing identity, to re-trigger
                                      # the permission prompts

Built with AudioTee

 talat — private, local-only meeting transcription for macOS. Captures system audio via AudioTee and runs real-time speech recognition, speaker diarization, and searchable notes entirely on-device. As featured in TechCrunch.

References / useful links

• Apple Core Audio Taps Documentation
• AudioCap Implementation
• AudioTee.js

License

The MIT License

Copyright (C) 2025 Nick Payne.