Stock Video Keyword Generator: Why Video Metadata Is Harder Than Photo Metadata — and How to Get It Right
Key Takeaways
Video metadata requires three distinct vocabulary layers that photo metadata doesn't: shot type terminology (establishing shot, b-roll, timelapse), motion descriptors (pan, pull-back, fly-through), and production use case keywords (background loop, lower third compatible, narrative opener)
Most AI keywording tools were built for photos and apply the same visual recognition approach to video — this produces keywords that describe the content of the frame but completely miss the shot-type and use-case vocabulary that video buyers search for
CyberStock handles both photo and video keyword generation from the same batch interface — the commercial relevance model applies to video content with the same buyer-intent training that makes it effective for photos
Video on Adobe Stock requires the same UTF-8 BOM CSV format as photos — the same compliance issues that cause manual cleanup on photo exports are present in video exports
Proper video metadata doubles the effective search coverage of a clip by capturing both what's in it and how buyers intend to use it
The Video Metadata Problem No One Talks About
When contributors think about stock video metadata, they usually think about the same process as stock photos: describe what's in the clip, add some related concepts, hit submit. This approach works adequately for photos, where the primary buyer search behavior is content-based ("business meeting," "coffee shop," "smiling woman"). For video, it captures only half of the relevant search behavior.
Video buyers search differently because they're not just looking for a subject — they're looking for a specific type of shot for a specific editorial or production purpose. A documentary editor searching for aerial footage of a city isn't searching "city aerial" — they're searching "city establishing shot" or "urban reveal footage" because they need a specific shot type for a specific structural moment in their edit. A motion graphics designer isn't searching "abstract background" — they're searching "seamless loop background" or "particle motion background" because they need footage with specific technical properties for their project.
This vocabulary layer — shot type, motion type, technical properties, editorial use case — is invisible to visual recognition tools because it's not visible in a single frame. It exists in the temporal and structural nature of the clip. The result is that most stock video metadata is systematically incomplete, and most contributors are leaving search coverage on the table for every clip they've submitted.
"A great drone shot of the Manhattan skyline with no shot-type keywords is invisible to the editor who needs a city establishing shot. A mediocre shot with 'urban aerial reveal, city establishing shot, Manhattan fly-through' in positions 1–5 gets the sale."
The Three Vocabulary Layers Every Video Clip Needs
Layer 1: Content Keywords (What's In It)
The foundational layer — the same commercial intent vocabulary that applies to photos. For a clip of a diverse business team in a conference room: business meeting, team collaboration, corporate diversity, professional discussion, meeting room, workplace culture, business professionals, inclusive team. These are buyer intent keywords, not literal descriptions ("people at table with laptop and whiteboard"). The same principles from photo keywording apply: use commercial phrases not object lists, put your strongest commercial terms in positions 1–10, avoid generic fillers.
Layer 2: Shot Type and Motion Keywords (How It Was Shot)
This is the layer unique to video. Every clip should be described by its structural and motion characteristics:
Shot type vocabulary: establishing shot, b-roll, cutaway, insert shot, POV shot, over-the-shoulder, two-shot, wide shot, medium shot, close-up, extreme close-up, aerial shot, drone shot
Motion vocabulary: pan, tilt, dolly in, dolly out, pull-back, push-in, tracking shot, handheld, gimbal smooth, timelapse, hyperlapse, slow motion, match cut, fly-through, orbit, ascending, descending
Duration and loop properties: seamless loop, 4-second loop, 10-second clip, loopable background, infinite loop, short clip
A standard Adobe Stock CSV column for video keywords should allocate positions 1–10 for content commercial keywords, positions 11–25 for shot type and motion vocabulary, and positions 26–45 for technical and atmospheric descriptors. CyberStock's commercial relevance model generates keyword sets that include this shot-type vocabulary for video clips, because video shot terminology is heavily represented in buyer search data.
Layer 3: Production Use Case Keywords (Why a Buyer Wants It)
The most sophisticated and most underutilized layer. These keywords describe the specific production contexts where a buyer would license this clip:
Editorial use cases: documentary b-roll, news package footage, explainer video background, educational content, social media background
Commercial use cases: advertising background, product promo backdrop, corporate presentation, event highlight, marketing video
Technical use cases: motion background, overlay footage, lower third compatible, title safe area, green screen alternative, color grade friendly
A clip of city streets at night with smooth gimbal movement and good exposure latitude might be keyworded for use as: night city background loop, urban nightlife b-roll, city motion background, nighttime street documentary footage, and after dark corporate footage. Each of these phrases targets a different buyer with a different project — and none of them are generated by a tool that only describes what's in the frame.
Platform-Specific Video Metadata Requirements
Adobe Stock Video
Adobe Stock uses the same CSV-based submission system for video as for photos. The same UTF-8 BOM encoding requirement applies. The 45-keyword cap applies to video clips. The keyword priority weighting on positions 1–10 applies to video search ranking. One difference: Adobe Stock video submissions require a separate "Category" field (Motion > B-Roll, Motion > Time-Lapse, Motion > Aerial, etc.) that helps buyers filter by content type — this field is not required for photos.
Adobe Stock's video review team is smaller than their photo review team, which means review times for video are longer (typically 14–21 days versus 10–14 days for photos). Video approval rates are also slightly more variable because technical quality review is more rigorous — a photo with slight exposure issues might be approved, but a video clip with inconsistent exposure, visible encoding artifacts, or audio issues will be rejected.
Pond5 Video
Pond5 has a richer metadata structure for video than any other platform. In addition to standard keywords and title, Pond5 requires: Category (Footage/Timelapse/Hyperlapse/etc.), collection assignment, contributor price setting, media description, and optional clip reel placement. The contributor price control is the defining difference — on Pond5, you set the price, and the platform takes their commission (40–60% depending on exclusivity). This means keyword quality directly affects your visibility at your chosen price point: a well-keyworded clip at $49 outperforms a poorly-keyworded clip at $25 in search results.
Shutterstock Video
Shutterstock's video metadata requirements are similar to photos: title, keywords (no cap but spam detection active), description. The key difference is that Shutterstock weights the title and description more heavily for video than Adobe Stock does. A title that accurately describes the clip's primary commercial use case ("Aerial Drone Timelapse of Urban City at Golden Hour") functions as an extended keyword field on Shutterstock and should include the most important commercial terms in natural language format.
How CyberStock Handles Video Metadata
CyberStock's batch keywording interface processes video clips by analyzing frame content and applying the same commercial intent model used for photos — trained on 50 million buyer searches that include video-specific search behavior. The output for a drone hyperlapse of an urban environment might be:
Positions 1–5: city hyperlapse, urban time-lapse aerial, drone city motion, metropolitan speed ramp, urban aerial fly-through
Positions 6–15: establishing shot, aerial b-roll, city motion background, downtown aerial, urban skyline reveal
Positions 16–30: smooth motion, gimbal stabilized, production quality, 4K aerial, city infrastructure, commercial use, loopable aerial, real estate aerial, documentary b-roll
This structure covers all three vocabulary layers in a single AI-generated output, requiring only a spot-check edit for any location-specific terms that the image analysis missed.
Generate complete 3-layer video keywords in seconds: cyberstock.lol — same batch interface for photos and video clips.