Obsidian Vault Performance Fix: Stop Search Hangs
Archived English guide, drafted with AI assistance and checked before publishing (editorial policy). New writing is in Korean on the Knowhow Seller log.
Obsidian vaults don't slow down because there are too many notes — plain markdown is tiny. They slow down because non-markdown files pile up inside the vault folder, and Obsidian's indexer has to scan all of them every time it builds or rebuilds a search index. The fix is usually not a plugin or a settings toggle; it's moving the offending files out.
What Causes Obsidian's Full-Vault Search to Hang or Freeze?
Obsidian's search and file indexer scan every file inside the vault folder, not just markdown notes. When the vault holds large binary files — video, design source files, spreadsheets, email exports — the indexer reads and hashes far more data than it would for plain text, so search stalls or the whole app appears to freeze during that scan.
- Quick Switcher or full-text search takes several seconds to return results, or the interface stops responding
- CPU or disk activity spikes right after the vault opens, after a sync, or after adding new attachments
- The slowdown tracks with new attachments, not new notes — a week of writing markdown doesn't change it, but dropping in a handful of PDFs or videos does
Which File Types Are Most Likely to Slow Down an Obsidian Vault?
Any file that's large relative to a typical note is a suspect: video, uncompressed images, design source files, spreadsheets, presentation decks, and email exports. Obsidian's indexer and file-system watcher touch these files alongside markdown even though it can't search their contents as text, so they add scan time without adding searchable value.
| File type | Typical size in a vault | Why it hurts performance | Recommended action |
|---|---|---|---|
| .md notes | A few KB each | Plain text — indexes almost instantly | Keep in the vault |
| .png / .jpg images | 100 KB–5 MB | Re-read for thumbnails and embeds on render | Keep small ones, compress or move large ones |
| .jfif images | 1–10 MB, common in camera/screenshot exports | Same as above, but frequently oversized and rarely compressed | Compress or move out |
| 1–50+ MB | Cached for the preview pane; large PDFs slow the file watcher | Move out, link externally | |
| .pptx / .xlsx / .xls | 1–100+ MB | Zipped office containers that are slow to hash at scale | Move out of the vault |
| .eml email exports | 1–20 MB, often with inline attachments | Rarely needed for note search; frequently duplicated | Move out or archive |
| .ai / .psd design files | 10–500+ MB | Large binary blobs the indexer still has to touch | Always move out |
| .mp4 / video | 50 MB–several GB | Single largest contributor to both size and scan time | Never store in the vault; link to external storage |
How Much Can Removing Binary Files Actually Shrink a Vault?
In a working knowledge-management vault used to run a daily content pipeline, moving 30 non-markdown files out dropped the vault from 115MB to 10.6MB — roughly a 91% reduction — and stopped full-vault search from hanging on that same handful of file types.
The files weren't exotic. They were the ordinary clutter that accumulates in any long-running vault: .xlsx and .xls spreadsheets, .eml email exports, .jfif screenshots, .pptx decks, and a few .ai design files. None of them needed to be searchable as text, and none of them needed to live inside the markdown vault to stay linked to the right notes — a plain reference note with a file path or external link did the job just as well.
Two details mattered in that cleanup:
- Files were moved, not deleted, into a folder outside the vault, so nothing was lost if a link needed to be restored
- Anything flagged as sensitive or private was moved without being opened, since the goal was size reduction, not review
This pattern is also widely reported by Obsidian's own community: users troubleshooting slow startup and search on the official forum consistently trace it back to large attachments — images, PDFs, audio, and other binary files — rather than note count (Obsidian Forum, forum.obsidian.md). A markdown note typically runs a few kilobytes; a single unresized screenshot or exported spreadsheet can outweigh hundreds of notes on its own.
How Do You Find the Large Non-Markdown Files Hurting Search?
The fastest diagnostic isn't opening files one by one — it's sorting the vault folder by size in your OS file manager (or a terminal command) and scanning for non-.md extensions above a few megabytes. Those large, non-text files are almost always the search-hang cause, not the note count.
- Windows: open the vault folder in File Explorer, switch to Details view, and sort by Size
- Mac: open the vault in Finder, use List view, and sort by Size
- Either OS: a quick terminal command (
duon Mac/Linux,Get-ChildItem | Sort-Object Lengthin PowerShell) lists the biggest files directly
How Do You Move Binary Files Out Without Breaking Links?
Move the files to a folder outside the vault first, then check Obsidian's Files and Links panel or a broken-link plugin for anything that still points at the old vault-relative path. Fix those references before deleting anything, and keep the moved files somewhere until you've confirmed nothing broke.
- Back up the vault (or confirm it's already version-controlled or synced) before moving anything
- Create one folder outside the vault for archived attachments
- Move flagged files there in batches, not all at once, so you can trace any broken link back to its cause
- Check Obsidian for broken embeds and either re-link to the new external path or remove the dead reference
- Keep the archive folder for a few weeks before deleting anything permanently
How Do You Keep the Vault From Bloating Again?
Treat this as a standing habit, not a one-time cleanup: attach large media by external reference instead of embedding the file itself, and re-sort the vault by file size on a regular schedule so new binaries get caught before they accumulate into another 100MB+ vault.
- Link out to cloud storage or a dedicated media folder instead of dragging large files into the vault
- Treat any inbox or scratch folder inside the vault as temporary — clear it on a schedule
- Re-run the size sort from the diagnostic step above monthly, not just when search starts hanging
Is Excluding Files From Search Indexing Enough, or Do You Need to Move Them?
Excluding files or folders from search in Obsidian's settings helps once a vault is already clean, but it doesn't shrink the vault, doesn't speed up sync, and doesn't stop file-system-level scans during startup. Moving heavy binaries out fixes the underlying cause; excluding them from search only hides the symptom inside the app.
Frequently Asked Questions
Does moving files out of the vault break my notes or links?
Not if you move files in batches and check Obsidian's broken-link warnings afterward — links that pointed at the old vault-relative path will show as broken until you re-link them to the new external location or remove the reference.
What file types should never go inside an Obsidian vault?
Video files, design source files like .ai or .psd, and any file regularly over 20-50MB are the safest to keep out entirely; link to them from a note instead of embedding them in the vault folder.
Will a plugin fix slow search instead of moving files?
Some plugins can exclude files from indexing or clean up unlinked attachments, which helps after cleanup, but no plugin removes the underlying files from disk — for a vault already bloated with binaries, moving the files is the actual fix.
How big can an Obsidian vault get before performance suffers?
There's no fixed limit — a vault of thousands of markdown notes stays fast because each note is a few kilobytes, but a vault a fraction of that note count can slow down fast once it holds gigabytes of video, PDFs, or design files.
Does deleting old daily notes help as much as removing binaries?
Usually not — markdown notes are so small that even years of daily notes rarely add up to the size of a handful of large attachments, so pruning notes has far less impact on vault size and search speed than removing binaries.