Real output from Blenda - not mockups. Every clip on this page was detected, masked, and re-encoded on a single workstation, with nothing uploaded anywhere.
Clips play on click and stream only when you ask for them, so the page stays light no matter how many we add. Each one is a finished export: the redaction is rendered into the pixels and the audio, exactly as it leaves the application.
A nine-second handheld clip of a packed street. Pick the detection classes - heads, people, vehicles, license plates, ID cards or screens - press Start Detections, and the AI makes a single pass. Every subject comes back as its own tracked redaction you can inspect, adjust or delete before anything is exported.
Not every clip calls for masking everyone. Here a single head is marked, labelled Suspect, and tracked through the shot while the officer stays visible. Then the audio: Blenda transcribes the clip on your own machine, you pick the words that identify someone, and only those words are bleeped in the exported track - the word, not the whole sentence.
Heads and identifying detail found and masked automatically, then rendered to a finished clip. Detections are tracked through the whole shot, so a subject who turns away or walks out of frame stays covered when they come back.
Switch it on and every pixel is masked by default. Then mark only what should stay visible - the subject of an interview, an officer's hands, a document on the table. Everything you don't approve stays blurred.
