Videntifier can process and identify tens of thousands of hours of video per day, while remembering the visual appearance of hundreds of thousands of hours of video in its database. On modest hardware too; The Videntifier front-end runs well on small devices, and the queries are low-bandwidth. The backend runs on commodity servers, and a bog-standard 1U machine can service thousands of users at the same time. The technology scales almost perfectly - both horizontally and vertically - so Videntifier can meet any of your organization's content identification needs.



Our identification is fine-grained, fast, and accurate. Compared to audio-based or "global" one-signature-per-frame identification, Videntifier just works. Like magic, it will identify almost anything you can throw at it – even cropped fragments of video – and tell you where exactly on the timeline you are.

We can do this because of our breakthrough database technology which allows us to use fine-grained and highly accurate visual interest points and still maintain world-leading search speeds.



Within milliseconds, the Videntifier database reliably identifies a heavily modified video frame against a database of 100.000+ hours of video. Videntifier also works live: We can ingest hundreds of channels of live TV simultaneously, and then in near-realtime identify the content appearing on devices or streaming platforms. Because the technology is highly efficient, we can do this on modest hardware, both on the backend and on-device. These features allows for both totally new user experiences and new forms of analytics.

Screen recording, subtitles, and compression



Videntifier easily recognizes modified material. This is due to the fine-grained local interest points we are able to use because of our proprietary database technology. Just to name some examples, Videntifier easily tolerates colour changes, rotation, mirroring, subtitles, compression, screen captures, and so on and so forth. Here are some samples:


Crop, rotation, and mirroring



Videntifier offers the entire pipeline from content ingestion to identification and action. On the client side, our technology grabs individual frames from streams, video files, or images, extracts the visual signature, and sends it one-way encrypted to our servers where they are either ingested or identified. The content cannot be recreated from the information that is sent over the network, which ensures privacy and data security.



By virtue of our proprietary database structures, Videntifier operates on modest and energy-efficient hardware. Our resource efficiency and scalability enables us to create visual connections on a much larger scale than previously thought possible. This scale means that radically new services and solutions have become economically and technologically feasible.