Deepware Scanner

What Deepware Scanner is

Deepware Scanner is a deepfake detector focused on face manipulation in video, which makes it synthetic-media forensics rather than text detection. The official site timed out during this check, so this description draws on third-party directories.

Video forensics addresses a different and in some ways harder problem than text. A manipulated clip of a public figure circulates faster than any correction, and the people most likely to share it are the least likely to question it.

What it does

You upload a video, or use the interface or offline software development kit, and it analyses frames for swapped or synthesised faces and other signs of manipulation. The loop is upload, scan, report.

Offering an offline kit is a meaningful option for organisations that cannot upload material, which is common in journalism and investigations where the source of a clip is protected. It also means the tool can be run repeatedly on material without creating a copy on a vendor's servers.

It is a video deepfake detector rather than a text checker, so the fit with this category is a mismatch.

Who it is for

It is aimed at media organisations, platforms and researchers vetting suspicious clips. Newsrooms verifying video before publication are the natural users.

What to keep in mind

Treat the result as one input in a verification process rather than as a verdict. Face-manipulation detection works well on some generation methods and poorly on others, and compression, low resolution and re-encoding degrade it further, which describes most of the video that actually circulates on social platforms. Provenance questions about who supplied the clip and whether earlier copies exist are often more decisive than the scan.

Because the input is video of real people, consent and retention matter. Video of a private individual obtained without their knowledge, held on a verification platform, is personal data with specific protections in several jurisdictions, and the practical questions are who can submit, what is retained, and how deletion works.

The official site was unreadable here and no privacy or certification text was visible in what could be read, so verify access and terms before uploading sensitive footage, and prefer an offline deployment if the material is confidential.

Two practical points. Test the detector on clips where the provenance is known, both authentic and manipulated, because that is the only way to learn its limits. And document the verification steps taken, since a newsroom's defence of a decision rests on the process rather than on a score. Note who ran the check and when, which is the first thing anyone reviewing the story will ask. A verification note of two lines is worth more later than a detailed memory of the process.