The reply starts.
Oreja reads the app's output level and acts on the first sound. Apple Music stops in pause mode, or drops to the level you set in lower-volume mode.
mode · pause, or lower volume
Menu bar app for macOS 14.2 and later
Oreja pauses Apple Music while ChatGPT speaks, or lowers it to a level you pick, then puts it back where you left it. It follows the app's audio level, so it works with any voice in any language and never listens to the words.
Vincent gave away one ear. The other one does all the listening.
The handoff
Oreja reads the app's output level and acts on the first sound. Apple Music stops in pause mode, or drops to the level you set in lower-volume mode.
mode · pause, or lower volume
Music stays down through the conversation hold, so the next reply does not restart it in the middle of your sentence.
conversation hold · 12 seconds, adjustable
After the quiet interval, playback returns at the level you chose and the same second of the same track it left.
resume · 100%, same position
Keep scrolling
The interface
The menu bar holds the panel: one status, one switch, the music line. Everything you decide lives in a single Settings window — the app to follow, what happens while ChatGPT speaks, the level music drops to, and when it comes back.
Settings
Oreja only touches playback while the selected AI source is speaking.
Behavior
AI source
Privacy
That is the whole input: one level from macOS's process tap, held in memory for a moment and thrown away. Oreja keeps a single answer, speaking or quiet, and nothing else.
Samples exist in memory for milliseconds. None are written to disk and none leave your Mac.
There is no account to create and no analytics or crash reporting in the bundle. The app contains no network code.
Pause Music yourself and Oreja leaves it alone. Resume during a reply and Oreja lets go of its claim.
Oreja starts no playback and changes no queue. Volume moves only in lower-volume mode.
Questions
No. It reads the output level of the selected application's audio and keeps only an active or quiet state. No samples are stored, no speech is recognised, and the app contains no networking code.
No. Oreja does not talk to OpenAI. It observes the local audio output of the app you choose and controls Apple Music through macOS automation.
Apple Music is the one player Oreja controls today. More players are coming.
Both. Pause mode stops Music while the AI speaks and resumes it afterward. Lower-volume mode keeps Music playing quietly and restores the level you set when the reply ends.
Not a normal one. Oreja listens for the shape of speech — the loudness rising and falling with every syllable — so a short blip, a steady alert tone or a chime that fades once is ignored, and Diagnostics tells you when that happened. A melody-style ringtone can still pass, and browser sources capture every tab in that browser, so a dedicated window or the desktop app gives the cleanest signal.
Your choice wins. If you pause Music, Oreja will not resume it. If you resume Music while ChatGPT is speaking, Oreja releases its claim and waits for the next activity before acting again.
Quitting while Oreja owns a pause resumes the music first. After a crash, the pause journal restores playback on the next launch only when the same track is still paused and the pause is recent. If the track changed or Music quit, Oreja leaves it alone.
It is ad-hoc signed and not notarized, so another Mac may ask for a one-time Gatekeeper approval. It is an independent project, not affiliated with, endorsed by or supported by OpenAI or Apple.
Morning
Download the 0.2.1 build for macOS 14.2 or later. It is ad-hoc signed and not notarized, so macOS asks for approval once. The source builds with one script and no dependencies.