Thrum

What is localspeechrecognition, and why is it using CPU?

Last reviewed: 19 November 2026. Path and bundle checked on macOS 26.3.1.

localspeechrecognition is a helper inside Apple's Speech framework, which apps use to turn spoken audio into text. As its name says, it handles local speech recognition, done on the Mac rather than on a server, so it uses CPU and memory while an app transcribes speech. It stops when the app finishes or is closed.

localspeechrecognition at a glance

localspeechrecognition
Made byApple (part of macOS, Speech framework)
TypeXPC service
Location/System/Library/Frameworks/Speech.framework/Versions/A/XPCServices/localspeechrecognition.xpc/Contents/MacOS/localspeechrecognition
Runs asyour user
Bundle IDcom.apple.speech.localspeechrecognition
Can you quit it?Yes, but a transcription in progress in the app that uses it may stop.

What is localspeechrecognition?

Apps can recognize speech with Apple's Speech framework. A recognition request can be set to keep its audio on the device, in which case recognition is done locally instead of being sent over the network. localspeechrecognition is an XPC service (a helper started on demand) inside Speech.framework whose bundle identifier is com.apple.speech.localspeechrecognition.

It runs as your user, and only while an app is using local speech recognition. Listening for the Siri trigger phrase is a separate process, corespeechd.

Is localspeechrecognition safe?

Yes. localspeechrecognition is part of macOS, signed by Apple, and lives inside /System/Library/Frameworks/Speech.framework on the read-only system volume. Apps need your permission for speech recognition and the microphone.

Why does localspeechrecognition use so much CPU and memory?

How much CPU is normal for localspeechrecognition?

Not running, or near 0% CPU, when no app is transcribing. Steady CPU while an app transcribes is expected and stops when it finishes.

What can you do if localspeechrecognition keeps using CPU?

  1. Find the app. Quit apps that transcribe or caption audio one by one and watch localspeechrecognition in Activity Monitor.
  2. Review speech recognition access. Open System Settings > Privacy & Security > Speech Recognition and turn off apps that should not use it.
  3. Stop live captions you do not need. Turn off live transcription or captioning features in apps when you are not using them.

When should you worry about localspeechrecognition?

Look closer if it runs for long periods when you are not using any app that transcribes speech. Check Privacy & Security for apps with speech recognition access.

It should run only from inside Speech.framework.

How do you check localspeechrecognition in Activity Monitor and Terminal?

Open Activity Monitor (Applications > Utilities), and type localspeechrecognition in the search field. The CPU tab shows its % CPU. The Memory tab shows its Memory. Double-click the row for its parent process, memory details and open files.

In Terminal, the first command lists every localspeechrecognition process with its PID, user, CPU and memory (RSS is resident memory in kilobytes).

ps -axo pid,user,%cpu,%mem,rss,command | grep -F 'localspeechrecognition' | grep -v grep

How can Thrum help with localspeechrecognition?

Thrum keeps a live CPU graph in the menu bar. One click shows the eight processes using the most CPU, so you can see when localspeechrecognition is among them and how much it takes. An optional alert tells you when a process stays above a CPU level you choose (50 to 200% of one core) for 1 to 10 minutes. Click localspeechrecognition in the list for its PID, threads and memory, and Quit or Force Quit it after a confirmation.

Thrum is $4.99 once with a 14-day free trial and no account. It requires macOS 26 Tahoe on Apple silicon or Intel.

Questions about localspeechrecognition

Does localspeechrecognition send my voice to Apple?

Its job is recognition on the device. Apps choose whether a request must stay on the device; Apple's documentation says on-device requests do not send audio over the network.

Is it the same as Siri?

No. It serves apps using the Speech framework. Siri's voice trigger is handled by corespeechd.

Sources