CSV Metadata Templates: One Master File, Many Stock Agencies

Adobe wants a Title column, Shutterstock wants Description, Vecteezy wants both. Here is the master column set that exports cleanly to each one.

There is no universal CSV for stock agencies. Adobe Stock wants a column called Title, Shutterstock wants one called Description, Vecteezy asks for both and requires neither, and Dreamstime splits the job across two spreadsheets that must be uploaded in a particular order. The data underneath is nearly identical at every agency. The header names and the delivery route are not.

That gap is where bulk uploads quietly fail. A contributor builds a clean sheet for one agency, copies it for the next, and the second agency rejects the whole file because a header does not match character for character. Nothing in the metadata was wrong.

The fix is not a single magic template, because one does not exist. It is one master sheet carrying a superset of fields, plus a short map of which agency renames what. Everything below comes from each agency's own published contributor documentation rather than from forum summaries.

The fields every agency asks for

Strip the agencies down and three fields carry almost all the weight: the filename that ties a row to a file, a human sentence describing the image, and a comma-separated keyword list. Adobe Stock makes all three required. Shutterstock makes all three required and adds a fourth, Categories. Vecteezy requires only the filename and treats title, description and keywords as optional.

The differences start with naming. Adobe's descriptive field is Title, capped at 70 characters, plain text, no commas. Shutterstock's equivalent is Description. Vecteezy asks for Filename, Title, Description and Keywords as its first four columns, in that order. One sentence of yours therefore lands in a column called Title at one agency and Description at another, which is the single most common mapping mistake and the reason standardising titles and descriptions across agencies is worth doing once rather than per upload.

Master metadata CSV builder

A superset of the columns Adobe Stock, Shutterstock, Vecteezy and Dreamstime ask for, with Adobe's published ceilings applied: filename 30 characters, title 70 characters, 50 keywords maximum, and Shutterstock's advised 7 to 50 keyword range.

What each agency's CSV actually expects

Five agencies are worth comparing side by side, because four of them publish a real specification and the fifth deliberately does not. Where an agency states nothing, the table below says so instead of filling the gap with a plausible guess.

CSV columns and upload routes by agency

Adobe Stock, Shutterstock, Vecteezy, Dreamstime and Magnific compared on the descriptive column name, keyword handling, category field, yes or no flags, where the CSV is uploaded and which limits each agency publishes.

Adobe Stock: the agency with hard numbers

Adobe is the only one of the five that publishes exact ceilings, and its CSV requirements page is worth reading once in full. The filename cell is limited to 30 characters and must match the file exactly, extension included. Title is capped at 70 characters. Keywords max out at 50 and should be ordered by importance, since position carries weight in Adobe's search. Category takes a number rather than a word, and the optional Releases column takes the exact filename of a model or property release, which has to be a JPEG and is capped at 30 characters too.

The file itself is constrained as well: a genuine .csv saved as UTF-8, a maximum of 5,000 rows, and a maximum of 1 MB. Spreadsheet formats are rejected outright, so an .xlsx renamed to .csv will not pass. Adobe's guidance for non-US Excel installs is to save as CSV UTF-8, or to re-save through a plain text editor if that option is missing. That encoding step matters more than it looks, because it is what stops accented and non-English keywords from arriving as mangled characters.

Shutterstock: sequence matters more than format

Shutterstock's sheet runs to seven columns. Filename, Description, Keywords and Categories are required; Illustration, Mature Content and Editorial are optional yes or no flags. Categories takes one or two values from Shutterstock's own list. Its metadata submission guidance states that the column names must match exactly or the file is rejected, which is why the Shutterstock CSV template is better copied than retyped.

The part that catches people is sequencing. The files have to arrive before the CSV, or the rows resolve to nothing and the upload returns a file-not-found error. The CSV also goes up through a dedicated CSV button on the Submit page rather than over FTPS alongside the images, so a contributor who drops everything into one FTP batch watches the images appear with no metadata attached. Shutterstock warns separately that a large CSV carrying rows for assets outside the current submission processes slowly and can hold up the batch, which is an argument for trimming the sheet to each submission rather than keeping one growing file. The wider Shutterstock upload workflow covers where this sits relative to review.

Dreamstime: two sheets, in a fixed order

Dreamstime splits metadata across an image spreadsheet and a model release spreadsheet, each exported to its own CSV. Its CSV tutorial is explicit that the release sheet goes first, into the modelrelease subfolder on the FTP, because the image sheet references the release IDs Dreamstime assigns once it has processed that file. The image CSV then goes to the main FTP folder. Filenames in both sheets have to match the submitted files, and only files named in a sheet get populated at all. When exporting, Dreamstime asks for a comma as the field delimiter and a double quote as the text delimiter. If you handle releases and categorisation for several agencies, this is the one that will not tolerate a wrong order.

Vecteezy and Magnific: the looser end

Vecteezy keeps it to four columns, and says so plainly: Filename, Title, Description, Keywords, in that order, with only the filename required and keywords comma-separated. The filename rule hides a trap. For FTP uploads Vecteezy wants the extension left off; for Add Data uploads the extension is optional. Spaces in filenames should be avoided either way. The CSV travels with the source files on FTP, or goes up through an Upload CSV control on the Add Data page afterwards. The rest of the Vecteezy metadata requirements sit outside the sheet.

Magnific, which contributors still often call Freepik, publishes no column specification at all. Its FTP upload documentation points contributors to a CSV upload or direct keyword entry inside the contributor panel rather than inside the FTP transfer itself, and separately requires that photos arrive with titles and keywords already embedded in the JPEG. FTP access is gated at 500 published files. Without a published column list, the defensible approach is embedded metadata as the primary carrier and the panel for anything missing, which is how the Magnific upload workflow reads in practice.

Why not embed everything and skip the CSV?

Embedded IPTC and XMP is the better primary carrier wherever an agency reads it, because it travels inside the file and survives a re-upload. CSVs persist for two reasons. Not every agency resolves embedded fields the same way, and some fields have no embedded equivalent at all: Shutterstock's Illustration, Mature Content and Editorial flags, or Adobe's category number and release filename, are agency concepts rather than IPTC ones. In practice the two are complementary. Embed what is portable, carry the agency-specific flags in the sheet, and keep an eye on the export step where metadata tends to disappear.

Filenames break more uploads than keywords do

Three different rules apply to the same column. Adobe wants the exact filename including the extension, inside 30 characters. Vecteezy wants it without the extension over FTP. Dreamstime wants it to match whatever was submitted, in both sheets. A master sheet that stores one filename string therefore needs two derived columns, one with the extension and one without, rather than a manual pass before every export. It is also the reason a file naming scheme is worth settling before a library grows.

Length is the quieter failure. A 30-character ceiling disappears fast once a naming scheme carries a date, a shoot code and a sequence number, and the truncation does not announce itself. Decide the scheme against the strictest agency rather than the loosest, because renaming after submission does not propagate to anything already in review.

Building the master sheet

Hold one row per file and a superset of columns: filename, extension, title, description, keyword list, a category column per agency, release filename, and the three Shutterstock flags. Nothing in that set is agency-specific except the category numbers, which have to be mapped per agency because no two use the same taxonomy, and a wrong one is tedious to fix in bulk afterwards.

Each agency export then becomes a column selection and a rename, not a rewrite. Keep those exports as disposable derivatives rather than editing the master in place. The master is the copy you will come back to in a year when an agency changes a limit, when a new agency joins the list, or when you re-optimise an old portfolio and need the original keyword order back.

Two constraints are worth designing around rather than discovering. Keyword order carries weight where an agency says it does, Adobe among them, so the master should store keywords in deliberate priority order rather than alphabetically. And because ceilings differ, truncation should happen at export, cutting from the end of a priority-ordered list, rather than by hand or by padding a list out to the maximum, which is what gets a batch flagged for keyword stuffing.

Before you upload

Six checks catch nearly everything. Confirm the header row matches the target agency character for character. Confirm the file is a real CSV in UTF-8 rather than a renamed spreadsheet. Confirm the row count and file size sit inside the agency's stated limits. Confirm the filename column uses the extension convention that agency wants. Confirm the sheet holds rows only for the batch you are submitting. And confirm the files are already in place wherever the agency resolves rows against uploaded content.

Maintaining that by hand across several agencies is the point where it stops being a spreadsheet problem and becomes a workflow one. Rastock.ai holds one metadata record per file, applies each agency's field limits and banned or mandatory keyword rules at export time, and delivers over FTP to 10+ agencies with per-agency status, so the mapping lives in the tool instead of in a column you keep in your head. IPTC and CSV export stay open, nothing claims ownership of your files, there is no revenue share, and the trial runs 14 days without a card.

Frequently asked questions

Is there a single CSV template that works for every stock agency?

No. The underlying fields are nearly the same everywhere, but the header names, the filename convention and the upload route differ per agency, and several reject a file whose headers do not match exactly. What works is one master sheet holding a superset of fields, with a per-agency export that selects and renames columns rather than a single shared file.

Why does Adobe call it Title while Shutterstock calls it Description?

They are the same field in practice, a human sentence describing the image, but each agency names it its own way and expects that exact header in the CSV. Adobe caps Title at 70 characters and asks for plain text without commas. Shutterstock uses Description and publishes no length ceiling for it. Map the two deliberately rather than assuming one column serves both.

How many keywords should a stock metadata CSV carry?

Adobe Stock accepts a maximum of 50 and asks for them ordered by importance. Shutterstock's own best-practice page tells contributors to stay between 7 and 50. Because order matters where an agency says it does, store keywords in deliberate priority order in the master sheet and truncate from the end at export, rather than letting a spreadsheet cut them arbitrarily.

Can I upload the CSV together with the images over FTP?

It depends on the agency. Vecteezy accepts the CSV alongside the source files on FTP. Dreamstime expects two CSVs on FTP, the release sheet first. Shutterstock does not: its CSV goes through a dedicated button on the Submit page after the files have arrived, and sending it over FTPS instead leaves the images with no metadata attached.

Why does my CSV upload fail when the metadata looks correct?

The usual causes are structural rather than content-related: a header renamed or translated, a spreadsheet saved as .xlsx and renamed to .csv, the wrong text encoding, a row count or file size over the agency's ceiling, or filenames carrying an extension where that agency wants none. Check the header row and the file format before re-reading the metadata itself.

Should I keep the per-agency export files or only the master sheet?

Keep the master. The exports are disposable derivatives and regenerating one takes seconds. The master is what you need a year later when an agency changes a limit, when you re-keyword an old batch, or when you add a new agency, and it is the only copy holding every field, including ones a given agency never asked for.