Face replacement and roadmap

Choosing a Video Privacy Tool: A Practical Unseen Checklist

Choose a video privacy workflow around your footage: local or cloud processing, selective face masking, difficult shots, and a reviewed export.

On this page

A useful video privacy tool must do two jobs: handle your recording in a way you accept and hide the faces you intend to hide. Start with those requirements, then test a short section of your own footage.

In Unseen, the first choice is Local processing or Cloud processing. The second is whether to mask every detected face or keep selected people visible. Here is how to make those choices without committing a whole recording to a workflow you have not tested.

Decide whether the video can leave your device

Choose Local processing when the source should stay on your device. It is free, requires no account, and runs the video-processing workflow in your browser.

If you want to work disconnected, use Prepare for offline use, wait for readiness, and test import through export with harmless footage after disconnecting. Keep the prepared tab open. See how to check whether video is uploaded for the distinction between downloading engine assets and transferring media.

Choose Cloud processing when uploading is acceptable and you want remote GPU computing to handle the job. It requires sign-in and credits. Include upload and download time when evaluating speed; there is no fixed advantage for every recording.

Cloud videos and their results are kept in your gallery until you delete them. If sensitive material must not stay on a server, delete it right after downloading, or use local processing.

Match the selection to the scene

For a crowd where nobody should stay visible, Blur all avoids selecting individual entries.

For a presenter surrounded by bystanders, choose Select faces, then Select all, and mark the presenter's confirmed entries Keep. Leave the others Hide. Check for separate entries showing a profile or reflection.

For just one person, clear the selection and mark each of their appearances for treatment. The selective masking guide explains when this approach is worth the extra review.

Face swap is a different operation: it uses cloud processing and replaces all detected faces with a supplied reference photo. It is not the selective masking workflow.

Test the hardest thirty seconds

Include a head turn, an obstruction, a camera cut, and a background face if those occur in your recording. An easy front-facing shot will not tell you how the workflow handles your difficult moments.

Export the test and check:

If coverage is missing, changing Blur to Pixelate will not repair an absent mask. Try a more thorough local preset where appropriate. Persistent misses can be fixed in Review & fix faces: draw a box on the face, track it, and export again. A crop or a cut also works.

Check limits before preparing the full job

Local input has an approximately 2 GB file-size cap and no fixed duration cap, although device resources still matter. Ordinary cloud processing accepts up to 500 MB and ten minutes. Cloud face swap has a shorter four-minute duration limit.

For a long recording, trim unused material first and test a representative section before processing the rest.

Keep the reviewed export ready to publish

Open the downloaded file and check the complete result. Store that copy separately from untreated sources, and use it for thumbnails and shorter edits too.

Ready to test your footage? Open Unseen, select the appropriate processing mode, and start with the short section most likely to reveal a problem.

Try it on a short clip

Start with a difficult moment from your video, then inspect the exported result.

Open Unseen