Amazon Rekognition: the output, not the model
Our AI video ad for Amazon Rekognition, the AWS computer vision service, never shows a model or a line of code. Each use case becomes an output a buyer can judge: "SCHOOL BUS" and "BROADWAY BUS LANES" labelled on a live street shot, a red warning on one image in a list, "SHIRT Style 5361" on a clothes rack. See how we made it.
What works here:
- The video opens on the manual work: a clock, a pile of files, a loading bar.
- Every capability ends on a result the viewer can read, not a process they have to trust.
The move: List what your agent hands back at the end of a run, then build each scene around one of those results.
