Pond5 documents its keywording expectations across several contributor pages, and two of those pages disagree about the single most basic number: how many keywords a clip should carry. One gives a range of 5 to 50 with 10 to 20 described as optimal. Another says always include 40 to 50. Both are official, and neither carries a date.
The keyword count contradiction, stated plainly
It would be easier to pick the number that suits the argument and move on, but that is how incorrect figures spread through contributor forums. The honest position is that Pond5's file-preparation guidance and its metadata guidance point in different directions, the gap between 10 to 20 and 40 to 50 is large, and no revision date on either page tells you which is current.
There is a defensible way through it. Both pages agree the hard ceiling is 50, and the lower page's accepted range tops out there too, so a set in the low-to-mid 30s breaks neither rule: it sits inside the 5 to 50 range and lands close to the 40 to 50 target. More usefully, it keeps you honest. The argument for a smaller set is relevance, and the argument for a larger one is coverage — a set in the 30s gives coverage without the padding that a forced 50 always produces.
Titles: 80 characters, and three things that get stripped
The documented title limit is 80 characters, with roughly 40 to 80 given as the useful band. That is comfortable by stock standards and gives room for a subject, an action and a location without abbreviation.
Three specific things are named as unwanted, and all three come straight out of the way video files are produced. Numeric sequence codes such as 001, left over from camera naming. File extensions such as .mov, left over from the filename. And words repeated inside the same title, which happens when a title is assembled from a template rather than written. Each is a slot spent on something no buyer will ever type.
Title length limits across three agencies
Pond5 documents an 80-character title limit with around 40 to 80 as the useful band, Adobe Stock's own guidance asks for titles under 70 characters, and Shutterstock allows up to 200, so a title of 70 characters or fewer satisfies all three.
The practical consequence of that spread is worth noting if you distribute the same clip widely. Writing to the shortest limit in your set, rather than to each limit separately, removes an entire class of truncation problem. A title of 70 characters or fewer is valid everywhere in the table above.
The description has 500 characters and a different job
Five hundred characters is a lot of room, and the common mistake is to fill it by restating the title with adjectives. The description earns its keep when it carries the context a title cannot: what happens over the course of the clip, where it was shot, what is visible in the background, whether there is camera movement. That is information a buyer uses to decide whether to license, and none of it fits in 80 characters.
English is mandatory, with one narrow exception
Titles, descriptions and keywords all have to be in English. The single documented exception is a place name that is genuinely searched more often in its local form than in its English one, which may be used as it stands. That is a narrow allowance for geography, not an opening for bilingual keyword sets — adding your own language alongside English is still outside the rules.
If you shoot in a non-English market, the workable pattern is to write the English metadata properly and keep the local-language place name where it genuinely helps discovery. Resist the urge to translate the rest, because non-English metadata is named as unacceptable rather than merely discouraged.
Filler words, special characters and the slots they waste
Stunning and beautiful are called out by name. So are special characters. The underlying principle is the same one that governs every keyword decision: a term has value only if it excludes something. Stunning excludes nothing, because no buyer filters for it, and it occupies a slot that a location, an action or a technique could have taken.
Order a keyword set for a drone harbour clip
Drag a mixed keyword set into priority order for a stock video clip. Subject, action and location terms belong at the front, conceptual terms in the middle, and filler adjectives and file artefacts such as stunning or .mov do not belong in the set at all.
Ordering is not explicitly ranked by Pond5 the way it is by some agencies, but the discipline still pays. A set whose first ten terms describe the subject, the action and the place reads as deliberate to a reviewer and to a buyer scanning results, while a set that opens with mood words reads as padding even when every term is technically true.
Editorial captions follow a fixed template
Editorial footage is where improvisation gets a clip rejected, because the caption format is prescribed rather than suggested. The shape is city, then country with the state in brackets where one applies, then the date, then a factual description of what the footage shows. The month is spelled out as a word rather than written as a number.
The description half of an editorial caption carries its own constraint, and it is stricter than most contributors expect. It records what is visible and nothing else: no interpretation of motive, no characterisation of a crowd's mood, no promotional language. If a sentence could be argued with, it does not belong in an editorial caption.
Nothing enters review until the metadata is complete
This is the structural detail that trips up contributors arriving from photo agencies. Pond5's own explanation of how selling works states that keywords, a description, camera information and prices all have to be added before a file can be submitted, and that a file stays in a needs-edits state until you explicitly save and submit it for review. An uploaded clip is not a submitted clip.
FTP makes this sharper still. The transfer brings the media across and nothing else, so embedded IPTC keywords in the file do not populate the Pond5 fields. The metadata arrives separately, through the web form or a CSV whose mandatory columns are the original filename, the title and the keywords. The field-by-field requirements cover that side in full, including the optional columns and how prices are handled.
Near-duplicates are a keywording problem too
Pond5 discourages uploading many visually similar files and asks contributors to keep the strongest shots rather than the full take. The documentation goes further than politeness here: repeated violations may bring a warning, and continued problems may lead to the account being removed.
The keywording connection is direct. Twelve near-identical clips with identical metadata compete with each other in the same search, so the best of them ranks lower than it would have alone. Culling is not just a courtesy to the reviewer — it concentrates whatever search weight the batch has onto the file most likely to earn.
A working routine for a mixed video batch
Cull first, before any metadata is written, because metadata written for a clip you then delete is pure waste. Then group what survives by location and subject, since clips from one place share most of their keyword set. Write the title to 70 characters so it travels, use the 500-character description for what the title cannot hold, and build the keyword set to the low-to-mid 30s.
Then run one pass whose only job is removal: sequence codes, file extensions, repeated words, filler adjectives, special characters and any non-English term that is not a place name. That pass takes a couple of minutes per group and removes almost every documented metadata violation on the platform.
Where this leaves a multi-agency workflow
None of these rules is difficult. The difficulty is that every agency has a different version of them: a different keyword ceiling, a different title length, a different stance on whether embedded metadata travels with the file. Doing that mapping by hand across several platforms is where the hours go. Rastock.ai generates the titles, descriptions and keywords, applies each platform's banned and mandatory keyword policy, and delivers to more than ten agencies over FTP or SFTP with status tracking per agency. There is no revenue share, no ownership claim, and IPTC and CSV export stay open.
What to do next
If you have clips sitting in a needs-edits state, that is the cheapest place to start: they have already been transferred and are waiting only on metadata. If instead you are getting files bounced back, the metadata rules above are only half the picture — the documented rejection reasons cover the technical and content side that keywording cannot fix.
If the bottleneck is volume rather than knowledge — you know what the metadata should say and there is simply too much of it — a 14-day free trial with no card required is enough to put a real batch through and see whether the cost of the tooling is covered by the hours it returns.
A closing word on the contradiction this article opened with. It is a reminder that agency documentation is not a single coherent rulebook but a set of pages written at different times by different people, and that the undated ones are the most dangerous. Before you commit a large batch to any number quoted here, open the contributor pages yourself and check whether they still say the same thing.
Frequently asked questions
How many keywords does Pond5 want?
Its documentation gives two answers. One contributor page sets the accepted range at 5 to 50 keywords and describes 10 to 20 as optimal, while another states that you should always include 40 to 50. Both are official and neither is dated. A set in the 30s satisfies the lower page's range and comes close to the higher page's target, which is the practical compromise.
How long can a Pond5 title be?
The documented limit is 80 characters, with roughly 40 to 80 described as the useful range. That is longer than Adobe Stock, whose own guidance asks you to keep titles under 70 characters, and much shorter than Shutterstock's 200-character ceiling. If you write one title for several agencies, 70 characters or fewer satisfies all three.
Can I write Pond5 metadata in my own language?
No. English is required for titles, descriptions and keywords. The one documented exception is a place name that is genuinely searched more often in the local language than in English, which may be used in that form. Everything around it still has to be English, so this is not a route to bilingual keyword sets.
What is banned from a Pond5 title?
Three things are named specifically: numeric sequence codes such as 001, file extensions such as .mov, and words repeated within the same title. Filler adjectives like stunning and beautiful are also unwanted, as are special characters. Each of those slots is better spent on a subject, an action or a location term.
How should editorial clips be titled on Pond5?
Editorial captions follow a fixed shape: city, country with the state in brackets where it applies, then the date, then a factual description of what the footage shows. The month is written as a word rather than a number. The description stays factual with no interpretation, no opinion and no promotional language.
Does FTP upload carry my keywords to Pond5?
No, and this catches people out. FTP brings the media file across only. Titles, descriptions, keywords, camera information and prices are supplied separately through the web form or a CSV, and a file sits in a needs-edits state until that metadata is complete and you submit it for review. Embedded IPTC keywords do not substitute for that step.