Images & photos · Image Metadata Privacy Tool
IPTC IIM vs XMP: Why Photos Carry the Same Caption Twice
· Background
image-privacy file-formats image-editing
News photos carried captions and bylines in the IPTC Information Interchange Model long before Adobe's XMP re-expressed those fields in XML, and many files now carry both blocks without agreeing. This post explains the history and the synchronisation problem.
Two captions, one photo — the concrete confusion when the IPTC block and the XMP packet disagree
A JPEG can carry caption-like information in more than one metadata family, and those copies can disagree after different editors update different blocks. ToolAcre does not choose a winning caption. Its JPEG classifier recognises Photoshop/IPTC in APP13 and XMP in APP1 as removable metadata, preserving the original privately while making a clean delivery copy.
IPTC IIM as a newsroom metadata block
The workbook attached a specific year and newsroom history to IPTC IIM, but those facts are not established by repository sources. The implementation-level fact is enough for this workflow: Photoshop resource data with an IPTC payload is recognised by its APP13 signature and removed as a whole segment rather than parsed into editable caption fields.
How IIM got into JPEGs — Photoshop's File Info and the APP13 Photoshop image-resource block that carries IIM inside a JPEG
JPEG application segments let several metadata systems coexist before the compressed scan. APP13 can hold Photoshop/IPTC resources while APP1 can separately hold Exif or XMP. A tool that removes only APP1 Exif can therefore leave an APP13 caption behind, which is why this cleaner classifies each supported segment by purpose rather than using EXIF as a synonym for every metadata block.
XMP as a separate XML metadata packet
The workbook also supplied a year and vendor history for XMP that the repository does not verify. The grounded distinction is structural: XMP is detected as a separate packet in JPEG and WebP, reported as present in WebP, and removed without parsing its XML. This tool cannot compare XMP caption text with an IPTC value it never decodes.
The synchronisation problem — software that updates one block and not the other, and the Metadata Working Group guidelines for reconciling them
Synchronisation problems arise when two representations are edited independently, but determining which field is authoritative belongs to an asset-management workflow. ToolAcre avoids silently reconciling values. Preserve the archival master, resolve caption conflicts in software that understands both schemas and use whole-block stripping only on a derivative whose metadata no longer needs to travel.
Encoding pitfalls — IIM's Latin-1 default and the coded-character-set marker, versus XMP's UTF-8
Encoding differences can make legacy caption bytes difficult to interpret, yet the repository has no IPTC decoder to diagnose them. Removal is simpler and safer for the publishing copy because the APP13 segment is skipped without rewriting its contents. Do not use an empty EXIF field table as evidence that APP13 or XMP was absent; inspection and removal have deliberately different coverage.
What this does not cover — how to author or correct captions; this post is about reading and removing, not writing
This cleaner is not a caption authoring, newsroom archive or metadata reconciliation tool. It also keeps ICC and rendering markers, and it cannot remove text burned into pixels. If caption metadata is part of a delivery contract, stripping it is the wrong operation; if it is private production history, retain it only in the controlled master.
Takeaway: assume both blocks exist — the Image Metadata Privacy Tool inspects and strips IPTC and XMP together, so neither copy survives unnoticed
Assume parallel metadata blocks may exist until the segment report says otherwise. On a delivery derivative, remove APP13 Photoshop/IPTC, APP1 XMP and APP1 Exif together, verify the result and retain the source where captions can still be managed. The value is predictable block handling, not an invented merger between metadata standards.