macOS Sequoia storage

com.apple.mediaanalysisd Cache Eating Storage on Sequoia? The 64MB-Per-Hour Bug Fix

If your Mac's storage meter shows a swelling "System Data" or "Other" wedge after upgrading to macOS 15 Sequoia, and Activity Monitor points at a process called mediaanalysisd, you have likely hit the most widely reported storage runaway on Sequoia: a cache that can grow at roughly 64 MB per hour while Photos is indexing, and that Apple has only partially addressed through the 15.3.x patch series. This article covers exactly what com.apple.mediaanalysisd is, where it stores its data, how to measure and reclaim the space today, and which macOS settings reduce the odds of it happening again.

What Is mediaanalysisd?

mediaanalysisd is the on-device machine-learning daemon that powers several Photos features: face recognition, scene and object classification, "Memories" generation, and the semantic search that lets you type "beach 2023" and find matching photos without uploading anything to iCloud. It runs entirely on your Mac, which is the good news. The bad news is that its working cache, written to a sandboxed container under your home folder, has exhibited an incremental-file bug on Sequoia where the daemon writes a fresh analysis chunk roughly every hour and does not reliably prune older chunks.

The result is a container directory that can reach 20–40 GB or more on Macs with large Photos libraries, sometimes without the Photos app ever being in the foreground.

The Exact Path Where the Cache Lives

Open Finder, press Shift-Command-G, and paste this path:

~/Library/Containers/com.apple.mediaanalysisd

Inside you will find a Data subfolder structured like a standard sandboxed container. The bloat almost always lives under:

~/Library/Containers/com.apple.mediaanalysisd/Data/Library/Caches

You may also see significant bulk under:

~/Library/Containers/com.apple.mediaanalysisd/Data/Library/Application Support

To get the exact size without leaving Terminal:

du -sh ~/Library/Containers/com.apple.mediaanalysisd/Data/Library/Caches
du -sh ~/Library/Containers/com.apple.mediaanalysisd/Data/Library/Application\ Support

If either number is in the multi-gigabyte range, you are looking at the runaway cache. Anything under a few hundred megabytes is within normal bounds for an active library.

Does macOS 15.3.x Fix It?

Apple's release notes for Sequoia 15.3 and 15.3.1 mention stability improvements to Photos analysis, and many users report that the hourly file pile-up becomes less aggressive after updating. However, the fix is not universal. Macs with large libraries (tens of thousands of photos), or machines that were offline for extended periods before reconnecting, frequently resume the runaway behavior after the update because mediaanalysisd re-indexes on library changes.

The practical answer: update to the latest 15.x point release first, then verify the container size has not crept back up a few days later. Updating alone is not a reliable permanent fix for everyone.

How to Safely Reclaim the Space

The cache is regenerable. Deleting it does not delete your photos, faces, or albums. What you lose temporarily is the ML index, which means Photos will re-analyze your library after a clear. On a large library that re-analysis can take several hours and will warm your CPU and fans. Plan accordingly.

Step 1: Quit Photos and stop the daemon

osascript -e 'quit app "Photos"'
launchctl stop com.apple.mediaanalysisd

The launchctl stop command halts the daemon for the current session. It will restart automatically when Photos reopens or on your next login, so move quickly.

Step 2: Delete only the Caches subdirectory

Deleting only the Caches folder is the conservative approach. It leaves preferences and the core model database intact, which can shorten re-analysis time:

rm -rf ~/Library/Containers/com.apple.mediaanalysisd/Data/Library/Caches

If the Application Support folder is also very large, you can remove it the same way, but be aware that Photos will perform a deeper re-index pass:

rm -rf ~/Library/Containers/com.apple.mediaanalysisd/Data/Library/Application\ Support

Step 3: Re-open Photos and let it rebuild

Open Photos, plug in power if you are on a MacBook, and let the library process in the background. The progress bar under "Photos" in the sidebar shows when analysis is running. Avoid force-quitting Photos during this phase or the index can fragment again.

Reducing Storage Impact with Photos Settings

Several Photos settings affect how aggressively mediaanalysisd runs and how large its cache grows.

Disable "Enhanced Visual Search" (Sequoia addition)

Sequoia added on-device visual search using a larger neural model. On Macs without Apple silicon, this model requires more cache space per image. You can disable it in Photos > Settings > General by unchecking "Enhanced Visual Search". This does not affect standard face recognition or Memories.

Pause and resume analysis strategically

In Photos, hold the Option key and open the Photos menu. You will see a "Pause Analysis" option that is hidden in the normal menu. Pausing analysis stops mediaanalysisd from writing new cache files until you resume, which is useful if you need the disk space temporarily before a deadline.

Reduce iCloud sync frequency

Every batch of photos synced from iCloud triggers a new analysis pass. If you have a large iCloud library syncing to a Mac with limited storage, consider setting Photos > Settings > iCloud to "Optimize Mac Storage" rather than "Download Originals". Fewer originals on disk means less to analyze and a smaller cache footprint.

Monitoring the Cache Over Time

One cleared, keep an eye on the container size over the next 48 to 72 hours. A simple Terminal one-liner can be run anytime:

du -sh ~/Library/Containers/com.apple.mediaanalysisd

If you see the number climbing back past 5 GB within a day or two without a large library import, the bug is still active on your machine. At that point, filing a Feedback Assistant report (Command-Space, type "Feedback Assistant") with a sysdiagnose attached gives Apple the diagnostic data needed to address it in a future point release.

Other com.apple.mediaanalysisd Symptoms to Know

  • High CPU from mediaanalysisd after login: Normal for up to an hour after boot if Photos has pending analysis. Sustained high CPU for days points to a stuck re-index loop. Quitting Photos and running launchctl stop com.apple.mediaanalysisd usually breaks the loop.
  • com.apple.mediaanalysisd taking space alongside com.apple.photoanalysisd: These are two separate daemons. photoanalysisd handles face detection; mediaanalysisd handles the broader ML scene and object pass. Both can bloat independently. The same cache-clear approach applies to photoanalysisd at ~/Library/Containers/com.apple.photoanalysisd/Data/Library/Caches.
  • mediaanalysisd filling drive after waking from sleep: Reported on Macs that go to sleep mid-analysis. The daemon resumes and writes fresh chunks instead of continuing the existing file. This is the mechanism behind the "64MB per hour" pattern.

What About "Photos cache huge on Sequoia" Reports?

The phrase "Photos cache huge Sequoia" covers a cluster of related issues. Beyond mediaanalysisd, three other locations deserve a check if your storage is still unexpectedly large after clearing the daemon cache:

  • ~/Pictures/Photos Library.photoslibrary: Your actual library. Right-click it in Finder and choose Get Info to see its true size. If originals are set to download and you have a large iCloud library, this is expected.
  • ~/Library/Caches/com.apple.Photos: The Photos app's own UI cache. Smaller than the daemon container, but worth checking.
  • ~/Library/Application Support/com.apple.MediaLibraryService: Metadata used across Apple media apps. Rarely the culprit, but grows on Macs with large music or video libraries.

If you want a bird's-eye view of which containers are actually using the most space without opening half a dozen Terminal tabs, Crumb's disk map surfaces oversized container caches like mediaanalysisd by name, shows the exact size, and lets you clear them with a reviewed, reversible plan without touching the originals in your Photos library.

Reclaim your disk in one click

Crumb audits your whole Mac, tells you what's safe to delete, and frees the space in seconds — private, local, and Apple-notarized.

Download Crumb for macOS

Frequently asked questions

Is it safe to delete the com.apple.mediaanalysisd cache on Sequoia?
Yes. The cache is fully regenerable and contains only working data for on-device ML analysis, not your actual photos, faces, or albums. After deletion, Photos will rebuild the index the next time it is open, which can take a few hours on large libraries but causes no data loss.
Will clearing the mediaanalysisd cache remove my named faces or Memories in Photos?
Named faces (people you have manually labeled) are stored in the Photos library database itself, not in the daemon container, so they survive the cache clear. Memories and Smart Albums based on ML classification will be temporarily unavailable and will regenerate as analysis completes.
How do I stop mediaanalysisd from filling my drive again after I clear it?
Update to the latest macOS 15 point release, disable Enhanced Visual Search in Photos Settings if you do not use it, and avoid large iCloud library syncs while the Mac is unplugged or on battery. You can also pause Photos analysis temporarily via the Option-click Photos menu when you need the disk headroom.
Does this cache issue affect Macs on macOS Sonoma or only Sequoia?
The runaway hourly-chunk pattern was most widely reported starting with macOS 15 Sequoia, which introduced the larger Enhanced Visual Search model. Sonoma users can see normal mediaanalysisd cache growth, but the multi-gigabyte runaway is a Sequoia-era behavior tied to the new analysis pipeline.
What is the difference between com.apple.mediaanalysisd and com.apple.photoanalysisd?
Both are Photos ML daemons but handle different tasks. photoanalysisd focuses on face detection and recognition, while mediaanalysisd handles scene, object, and semantic classification as well as the Enhanced Visual Search index introduced in Sequoia. Each has its own sandboxed container and cache, and both can grow large independently on Macs with big libraries.