Metadata embedder

The problem it solves

Most contributors keep metadata in one place only: the spreadsheet they upload alongside a batch. That works until the file leaves that batch. Re-upload the same JPG to a second marketplace and the keywords are gone. Send a client a finished frame and it arrives anonymous. Come back to an archive folder two years later and the only clue about what a file was for is its filename.

Embedded XMP fixes that permanently. Lightroom, Bridge, Windows File Explorer, macOS Finder, most stock upload forms and most asset managers read the same standard fields, so a file that carries its own title and keyword list gets recognised everywhere instead of being re-keyworded from scratch.

There is a second, newer reason: AI disclosure. Adobe Stock and Shutterstock both expect generative work to be declared, and the reliable way to declare it is the IPTC digital source type field written into the file rather than a note in a spreadsheet column. This tool can stamp trainedAlgorithmicMedia across a whole batch in one pass, which is far safer than remembering to tick a box per upload.

How it works, step by step

1. 1. Add files

   Drop in JPG or PNG files — up to 1000 per batch. Nothing uploads: each file is read as bytes in your own browser, and the originals on disk are never modified.

2. 2. Bring metadata in, or type it

   Paste a title and keyword list manually, or load the CSV or JSON you already exported from the metadata studio as a sidecar. Rows are matched to files by filename, so a 300-file batch inherits its metadata in one action.

3. 3. Set the shared fields

   Creator, copyright notice and the AI-generated flag usually apply to the whole batch, so they are set once rather than per file. Titles are capped at 200 characters and keyword lists at 50 entries to stay inside what reviewers accept.

4. 4. Byte-level insertion

   For JPEG, an XMP APP1 segment is inserted directly after the SOI marker; existing XMP is replaced rather than duplicated. For PNG, an iTXt chunk carrying the same packet is written before IDAT. Image data is copied through untouched, so there is no re-encode and no quality loss.

5. 5. Download

   Take files individually or as a ZIP. The output is a normal JPG or PNG that any reader can open — the metadata is simply now inside it.

What it deliberately does not do

It does not re-compress, resize or colour-manage anything. It does not strip your camera's EXIF: shutter, aperture, lens and capture date survive, because reviewers occasionally check them and because they are useful to you later. It also does not write to video containers — MP4 and MOV metadata is a different specification, and for clips the CSV remains the route the marketplaces actually read.

Questions about this tool

Does embedding XMP change how the image looks?

No. The compressed image data is copied byte for byte and only a metadata segment is added, so the pixels are identical to the file you started with and there is no second round of JPEG compression.

Will Adobe Stock and Shutterstock read the embedded fields?

Both read embedded title and keyword fields on upload and will pre-fill their forms from them, which saves a lot of typing. Treat that as a head start rather than a replacement for the CSV: category and release information still belong in the spreadsheet or the web form.

How do I flag AI-generated work correctly?

Turn on the AI-generated switch. It writes the IPTC digital source type as trainedAlgorithmicMedia, the value both marketplaces look for, into every file in the batch alongside your normal title and keywords.

Can I embed metadata into videos here?

Not yet. Only JPG and PNG are supported, because those two containers can be edited safely in the browser without re-encoding. For MP4 and MOV clips, export a CSV from the studio instead.

Do my files get uploaded anywhere?

No. Reading bytes, building the XMP packet and writing the new file all happen in the browser tab. You can watch the network panel stay silent for the whole batch.