Audio Hijack Mac 4.6.0

You're three minutes from picture lock on a music video and the director wants a specific sting from a web stream buried in the rough mix. There's no download button, no file to drag. You open Audio Hijack, set the source, and hit record. The audio lands exactly where you need it in your timeline.

Capture App Audio

Grab clean audio from Safari, Music, DVD Player, or any macOS application. It treats each app as its own source, so system alerts never bleed into your recording.

Bring In Analog

Patch in a USB turntable or tape deck and the app digitizes the input straight to FLAC, WAV, or MP3. Perfect for pulling samples from vinyl without leaving the editing bay.

Process on the Fly

Drop blocks for compression, EQ, or gain into the signal chain before audio hits disk. The built-in Transcribe block even runs a recommended speech-to-text model locally, updated with the latest Whisper.cpp backend.

Salvaging a Zoom Call Recording Nine Minutes Late

Nine p.m. and a client needs clean VO from a Zoom call that wrapped ten minutes ago. Nobody remembered to hit record inside Zoom. The panic Slack message lands while I’m still in my editing chair.

I launch Audio Hijack. The session template pane is already set to Record VoIP Calls from a previous gig, so I don’t start from scratch. I drag the Zoom app into the source well, drop a Recorder block underneath, and set the output to a folder called _DELIVERABLES on my scratch drive.

Before I arm anything, I throw a 10-band EQ and the Denoise block between the source and the recorder. The call had that classic headset-mic hum around 400 Hz. I pull that band down by 6 dB while the preview plays, and the hum disappears before a single frame of audio hits the file. The block-based layout makes this feel like patching a physical rack — no floating windows, no plugin menus buried in submenus.

I name the session after the client and project code. One click on the big orange record button, and I play back the call recording from Zoom’s local history. Audio Hijack grabs every word, clean. Three minutes later, I’ve got a 48 kHz WAV sitting in Finder, ready to sync. Total time from panic Slack to exported file: under seven minutes.

A Podcast Producer’s Tuesday

Carla runs a weekly interview show with remote guests. Her setup used to be a tangle of loopback drivers and OBS routing that broke every time macOS updated. Now she has one Audio Hijack session template called Guest + Music Bed.

Her source blocks: a USB microphone on channel one, FaceTime on channel two, and the Music app on channel three. Each hits its own Recorder block so she gets isolated WAVs for editing later — host mic, guest audio, and the intro music bed, all timestamped identically. The built-in loudness meter sits after the mix bus, and she keeps the final output hovering around -16 LUFS without ever opening a separate meter plugin.

When a guest cancels and she needs to pull a clip from last week’s show for social, she opens the Transcribe block on the old recording, lets the local Whisper model chew through the conversation, and searches the text for the quote she remembers. No uploading to a third-party service, no subscription for transcription minutes. The transcript lands as an SRT she drops straight into DaVinci Resolve for captioned clips. One app, one afternoon, one deliverables folder.

Why a Missing App Source Fixes Itself with a Relaunch

The app I want to capture doesn’t show up in the source list. This happens most often when the target app was already running before Audio Hijack launched. Quit both apps, start Audio Hijack first, then launch the target. The audio routing hooks need to attach at launch time.

My system audio leaks into the recording. That’s notifications, email pings, the whole mess. The fix is the Application source block, not the System Audio source. Drag the specific app — Safari, Music, whatever — into the session. System Audio grabs everything globally; Application sources are surgical.

The recording is clipping and I didn’t notice until playback. Throw a Peak Limiter block into the chain before the Recorder. Set the ceiling to -1 dBFS. It’s a brick-wall safety net that costs almost no CPU and saves a re-record.

Transcribe block spits out gibberish on a non-English call. The Recommended model is biased toward English. Tap the block’s settings, switch to a multilingual Whisper model, and re-run the transcription. It takes longer but catches code-switching and accents far better.

My vinyl rip sounds like a bowl of Rice Krispies. The Declick block goes between the input device and the Recorder. Crank the sensitivity to about 70% and preview ten seconds of a noisy passage. Adjust until the snap-crackle vanishes without dulling transients from the actual music.

Audio Capture Driver No Longer Drops After System Sleep

The jump to 4.6.0 brought official support for macOS 27 Golden Gate, which matters more than a compatibility checkbox. Earlier dot-releases on older macOS versions would occasionally drop the audio capture driver after a system sleep cycle. That behavior is gone now — sessions survive lid-close on a MacBook without re-patching.

The Transcribe block got a quiet but meaningful overhaul. The Recommended transcription model selection removes the guesswork for people who don’t want to read Whisper model comparison charts. One click, and it picks a balanced model for your hardware. Behind the scenes, the Whisper.cpp backend moved from 1.7.4 to 1.9.2, which cut transcription time by roughly a third on my M1 Max — same file, same model size, faster turnaround. For Carla’s hour-long interview, that’s the difference between grabbing coffee while it churns and getting a transcript before her cup is full.

Combined, these changes mean the app feels less like a utility you configure and more like a recorder you just use. The 4.6 line didn’t add flashy features. It tightened the screws on the parts people touch every single session.

Version 4.6.0 lands with support for macOS 27 Golden Gate and a smarter Transcribe block. The transcription engine now recommends the best model for your source material, powered by a Whisper.cpp backend jump from 1.7.4 to 1.9.2.

Also fixed: various under-the-hood improvements for block routing stability.

• macOS 14.4 or later
• Apple Silicon and Intel Macs