Nextcloud Photo Fixer

Nextcloud Photos orders by EXIF DateTimeOriginal — so if your archive came from phones, cameras, WhatsApp, screenshots, and a Google Takeout, your gallery is a mess. The fix isn't the gallery. The fix is the metadata.

"Great project and nice learning experience! Many other users will find this useful. 🌟" — Nextcloud, on LinkedIn

What it does

• Walks a Nextcloud user's Photos tree on disk.
• Prefers Google Takeout's photoTakenTime.timestamp from supplemental .json sidecars, which is the most trustworthy source for phone backups.
• Falls back to EXIF DateTimeOriginal for files that have it.
• Writes a coherent pair — EXIF tag and filesystem mtime — so Nextcloud's indexer agrees with ls -l.
• Is idempotent: re-running on a clean archive is a no-op. You can cron it.
• Logs every decision to /var/log/nextcloud_photo_fix.log so you can audit what happened after the fact — and then occ files:scan to let Nextcloud catch up.

Usage

# install prerequisites
$ sudo apt install jq libimage-exiftool-perl

# point it at a user's photo tree
$ sudo ./fix_photo_dates.sh \
      "/path/to/nextcloud/data/username/files/Photos"

# then tell Nextcloud to reindex
$ sudo -u www-data php occ files:scan --path=username/files/Photos

Design notes

• Source hierarchy. Not all metadata is equal. Google Takeout's JSON beats in-file EXIF, which beats mtime, which beats the filename.
• Idempotence via a guard. Files whose EXIF and mtime already agree with the best-available source are skipped entirely — the script can run nightly without burning the SSD.
• One log, one decision per file. Every mutation is one line: path, chosen source, old value, new value. Debuggable by grep.

Why it matters

Most self-hosters solve this by manually curating. The tool automates the same judgement calls — which source to trust, when to leave a file alone — across tens of thousands of files without repeatedly rewriting the ones you already fixed.

→ Tool on GitHub