Stock video keyword generator with multi-frame analysis

For every clip we extract three representative frames (start, middle, end), run vision AI on each, and merge the keywords, which catches scene changes a single frame misses. Camera movement is a separate optional step: we measure how the picture moved across the clip in your browser, so the move that reaches your CSV is one we actually observed.

MP4, MOV, and M4V — the formats Adobe Stock and Getty accept. Longer clips are sampled at start, middle, and end.

Yes, when it's clear: Zoom In, Zoom Out, Panning, Tilt, Tracking Shot, and Aerial View, using each platform's exact term. If the movement is ambiguous we add no movement keyword rather than guess.

2 credits per clip, regardless of length.

Standard MP4 / H.264 is measured in any browser. ProRes, HEVC and cinema-grade H.264 have to be decoded by the browser to be measured, and Chrome and Firefox skip them — tag those clips in Safari if you want movement keywords, or leave the box unticked and name the move in your shoot notes, which always wins over the measurement anyway. Tagging itself works everywhere regardless; only the movement keyword depends on this.

Usually yes, and for a short single-action clip more frames add nothing. But for a long take that travels through a scene, an activity that turns into another, or a reveal at the end, tick "Complex action" and we sample five points instead of three, so the merged keywords describe the whole sequence rather than the parts three frames happened to land on. That costs 1 extra credit per clip.

30 per drop — lower than the 100-photo cap because we decode frames from every clip in your browser before anything is sent, which is heavier on the tab than resizing a photo. The cap is per drop, not per export: keep dropping batches and they stack in the same review screen and download as one combined CSV, and re-dropping a file already in the export is skipped rather than charged again.