Dreamstime will accept a technically flawless file and still leave it parked in your unfinished queue because one text field is wrong. The platform validates title length, keyword count, category assignments and description text before a reviewer ever opens the image, and the rules differ from Adobe Stock and Shutterstock in ways that make copy-pasted metadata fail quietly.
This is a field-by-field reference to Dreamstime metadata requirements as they stand in 2026: what each field accepts, where the hard limits sit, how the CSV and FTP path differs from the web form, and which habits carried over from other agencies will cost you submissions.
What Dreamstime validates before a human sees your file
Dreamstime runs two separate gates. The first is automated and purely structural: does the file meet the format and resolution spec, and does every required metadata field contain something valid. The second is human review, where a moderator judges commercial value, technical quality and whether your keywords actually describe what is in the frame.
Most contributors who complain about slow approvals are stuck at the first gate without realising it. A row that fails validation does not generate a rejection email the way a moderator decision does. It simply does not advance, which is why files can sit in the unfinished section for days while you wait for a verdict that was never coming.
The four fields that carry mandatory rules are title, description, keywords and categories. Everything else in the submission form is either optional or a licensing toggle.
Title: 5 to 250 characters, English only
Dreamstime accepts titles between 5 and 250 characters. That upper bound is generous by microstock standards, and the temptation is to fill it. Do not. The title is what a buyer reads in search results and what Dreamstime's own search engine weights most heavily, so a 240-character title padded with synonyms reads as spam to both.
Dreamstime also requires that all submitted text is written in English. This trips up contributors working in other languages who translate keywords but leave a native-language title in place from their camera or cataloguing software. The row fails, and the failure reason is not always obvious from the interface.
Two practical rules hold up well here. Write the title as a plain descriptive sentence fragment a person would actually say out loud, and make it unique per file. Dreamstime explicitly asks contributors to differentiate near-identical images so similar frames reach different buyer searches instead of competing against each other.
Description: required, and not a second copy of the title
The description field is mandatory. A blank description will fail validation on the web form and will kill the row in a CSV batch, which is the single most common reason a large FTP batch comes back half-populated.
The useful way to think about the description is that it carries the context the title has no room for: where the scene is, what the subject is doing, what the mood or use case is, whether anything in frame is a recognisable location or event. Dreamstime's guidance to contributors is that information should be objective and accurate rather than aspirational, and that similar images should carry differentiated descriptions.
Dreamstime has also added an AI keywords tool inside the upload section that can produce a detailed description once you have entered the title. It is a reasonable starting point, but it describes what it sees and nothing else, so anything a viewer could not infer from the pixels alone still has to come from you.
Keywords: seven is the floor, not the target
Dreamstime requires a minimum of seven keywords per file. Rows carrying fewer than seven fail validation outright. That is a floor, not a recommendation, and seven keywords will not carry an image in a library of this size.
Dreamstime's reviewers check keywords against the image, and irrelevant tags are grounds for a bounce rather than a silent trim. This is the meaningful difference from agencies that simply ignore keywords they do not recognise: on Dreamstime, padding your list with aspirational concepts is an active risk, not a free bet.
A keyword set that survives review usually covers five layers: the literal subject, the action or state, the setting, the visible attributes such as colour and lighting, and the concept a buyer would search when they need this frame for a project. If a term does not fit one of those five, it probably does not belong on the file.
Also worth remembering: keywords embedded in the file's IPTC fields are read on upload, so whatever your cataloguing software wrote into the image travels with it. If you have inherited a messy legacy library, that embedded data is what Dreamstime sees first.
Categories: up to three, ordered by relevance
Dreamstime's submission structure allows up to three category assignments per file, and the CSV template reflects this with separate Category 1, Category 2 and Category 3 columns. Categories are required for the file to be findable, not optional metadata you can fill in later.
Order matters more than most contributors assume. The first category is the primary classification and should be the one a buyer browsing the taxonomy would expect. The second and third are for genuine secondary readings of the image, not for coverage. Assigning three loosely related categories to widen reach tends to weaken the placement of all three.
File and technical requirements that sit alongside the metadata
Metadata validation runs in parallel with the file spec, and a file that fails either one does not advance. Dreamstime's published requirements for images are:
- JPG format only, in RGB or sRGB colour space
- A minimum of 3 megapixels and a maximum of 70 megapixels
- Files free of visible noise, compression artefacts and over-processing
Separate spreadsheet templates exist for images, model releases, video and audio, so video and audio submissions follow their own field sets rather than the image one.
The resolution ceiling is the one that catches people out. Contributors shooting on high-megapixel bodies or upscaling AI output can push past 70 MP without noticing, and the file is refused at ingest rather than at review.
The CSV and FTP path: field order and processing sequence
For anything above a handful of files, Dreamstime's CSV route is the only workable option, and it has a strict sequence that is easy to get wrong. Dreamstime provides an Image spreadsheet template containing two sheets — an Image data sheet where you enter values, and an Image legend sheet that defines what each numeric code means. Licensing options are entered as numbers, not words: to enable the web extended licence on a row, you put 1 in the W-EL column.
The image template's columns cover Filename, Image Name, Description, Category 1, Category 2, Category 3, keywords, Free, W-EL, P-EL, SR-EL, SR-Price, Editorial, MR doc Ids and Pr Docs. Everything a reviewer sees and every licensing decision you are making lives in that single row.
The sequence Dreamstime documents is specific, and each step depends on the one before it:
- If you have model releases, upload the MR spreadsheet first and collect the Dreamstime-assigned MR ID numbers from the MR Library — those IDs are what you reference in the image sheet
- Upload the image files themselves via FTP and let them finish processing, so they appear in the unfinished section
- Convert the image spreadsheet to CSV and upload it into your FTP folder
- Only the filenames listed in the spreadsheet are auto-populated, so any mismatch between the file on the server and the string in the Filename column silently leaves that image blank
When you export the CSV, the field delimiter must be a comma and the text delimiter a double quote. Spreadsheet editors that default to a semicolon will produce a file Dreamstime cannot parse, and the batch will appear to do nothing at all.
Two exports are available and worth using rather than maintaining parallel records by hand: the MR Library exports a CSV of your model releases, and Upload History exports a CSV of what you have submitted. Reconciling against those two files is faster than auditing the platform by eye.
The filename column is also the reason a consistent naming convention pays off across agencies. If your originals, your CSV rows and your FTP directory all agree on a single canonical filename, this step stops being a source of errors entirely.
Where Dreamstime differs from Adobe Stock and Shutterstock
If you distribute the same shoot to several agencies, four differences matter enough to break a batch:
- Title length. Dreamstime's 250-character ceiling is far more permissive than the tight, sentence-length titles other agencies expect, so a set written for Dreamstime may need shortening elsewhere rather than the reverse.
- Mandatory description. Dreamstime requires it. Agencies that treat the description as optional will accept a batch that Dreamstime rejects row by row.
- Keyword floor. Seven is enforced, and reviewers actively check relevance rather than quietly discarding unrecognised terms.
- Three-category structure with a coded CSV legend. Numeric legend values for licensing toggles have no equivalent in a Shutterstock or Adobe Stock CSV, so a template built for one agency will not port.
This is the practical argument for keeping one canonical metadata record per file and generating each agency's export from it, rather than maintaining separate spreadsheets per platform. The moment those spreadsheets drift apart, you are debugging the difference instead of shooting.
Rastock AI is built around that model: one metadata record per file, agency-specific rules applied at export, and FTP delivery to 10+ agencies including Dreamstime from the same batch. Policy-aware tagging catches banned and mandatory terms before they reach a reviewer, and rollback lets you undo a bad batch rather than re-editing it file by file.
A pre-upload checklist
Run this before a batch leaves your machine, not after it stalls:
- Every title between 5 and 250 characters, in English, unique per file
- Every description populated — no blanks, no duplicates of the title
- At least seven keywords per file, every one of them defensible against the image
- Category 1 assigned deliberately; categories 2 and 3 only where genuinely relevant
- Files JPG, RGB or sRGB, between 3 and 70 megapixels
- CSV filenames matching the files on the FTP server exactly, character for character
- CSV saved with comma field delimiter and double-quote text delimiter
- MR spreadsheet uploaded and MR IDs collected before the image spreadsheet goes up
None of this is difficult. It is just unforgiving at volume, which is exactly where a single wrong delimiter or a missing description column turns a 400-file batch into an afternoon of manual repair. You can test the whole pipeline on a real batch during the 14-day free trial — no card required.
Frequently asked questions
How many keywords does Dreamstime require?
The documented minimum is seven keywords per file, and rows carrying fewer than seven fail validation before review. Seven is a floor rather than a target, and Dreamstime's reviewers check keywords against the image itself, so irrelevant terms are grounds for a bounce rather than being quietly ignored. A defensible set usually covers the literal subject, the action, the setting, visible attributes and the concept a buyer would search.
What is the title character limit on Dreamstime?
Titles must be between 5 and 250 characters. The ceiling is generous compared with other microstock agencies, but filling it works against you: the title is what buyers read in search results and what Dreamstime's search engine weights most heavily, so a long title padded with synonyms reads as spam. Dreamstime also asks that titles are unique per file rather than repeated across a series of similar frames.
Is a description required on Dreamstime?
Yes. The description field is mandatory, and a blank description fails validation on the web form and kills the row in a CSV batch. It is the most common reason a large FTP batch comes back only half-populated. Use it for context the title has no room for: location, what the subject is doing, mood, and use case. Dreamstime asks that similar images carry differentiated descriptions rather than a copy-pasted block.
Can I submit Dreamstime metadata in a language other than English?
No. Dreamstime requires that all submitted text, including the title, description and keywords, is written in English. This catches contributors who keyword in a local language for other agencies and translate only part of the record, or who leave a native-language title in place from their cataloguing software. Translate everything before submission, because editing metadata on files already in the queue means waiting through review again.
What file specifications do Dreamstime images need to meet?
Images must be JPG in RGB or sRGB colour space, with a minimum of 3 megapixels and a maximum of 70 megapixels, and free of noise and other artefacts. The upper limit is the one contributors hit unexpectedly, either from high-megapixel camera bodies or from upscaled AI output. Files outside the range are refused at ingest rather than at review, so they never generate a rejection notice you can act on.
How does the Dreamstime CSV upload work?
Dreamstime provides an image spreadsheet template with an Image data sheet and an Image legend sheet that defines numeric codes for licensing options. The sequence matters: upload the model release spreadsheet first and collect the assigned MR IDs, then upload the image files via FTP and let them process, then upload the CSV. Only filenames listed in the spreadsheet are auto-populated, and the export must use a comma field delimiter and a double-quote text delimiter.