How to Remove the Google Veo Watermark

By Calabi Labs Editorial Team · 2026-08-01

To handle a Google Veo watermark, crop out any visible logo, then run the file through Calabi to strip the invisible metadata layer platforms scan — the C2PA / Content Credentials manifest, XMP AI flags, generator tags, and encoder fingerprints — and inject a real phone-capture identity. Be clear-eyed: Veo also embeds an invisible pixel watermark that no metadata tool removes, and results vary by platform and source model.

What actually flags your file

A Google Veo export carries a layered provenance stack. The metadata layer includes a C2PA / Content Credentials manifest — a signed "made by AI" record stored as a JUMBF block that platforms read straight from the file. It survives cropping and re-uploading because it lives in the metadata, not the picture.

The container also holds XMP flags like DigitalSourceType: trainedAlgorithmicMedia, generator and tool tags, and encoder fingerprints such as Lavc or x264 SEI data. Those are invisible and machine-readable, and they're the signals most commonly used to identify AI video.

Veo goes a step further than most: it also embeds an invisible per-pixel watermark baked into the image itself. This is a different layer from metadata, and it's important to be honest about it — a re-encode can disrupt some patterns, but no tool can reliably remove an invisible pixel watermark, and results vary.

How Calabi handles it

Calabi is a one-pass file tool, not a video editor. It does not select, auto-detect, or paint over any region. For a visible Veo logo, the honest fix is a crop — Calabi does not erase logos pixel by pixel. What Calabi cleans is the metadata and provenance layer.

Stage one strips the signals: the C2PA/JUMBF manifest, the XMP trainedAlgorithmicMedia flag, generator tags, and encoder SEI fingerprints. Stage two injects an authentic phone-capture identity — Make, Model, Software version, GPS coordinates, capture timestamp, and a real-phone encoder name. Profiles include iPhone 17 Pro Max, iPhone 16 Pro, iPhone 15 Pro, Pixel 8 Pro, and Galaxy S24 Ultra. Stage three returns a forensic proof card so you can see the metadata result before download.

In verified runs, ExifTool showed 18 JUMBF/C2PA atoms reduced to 0 and 16 C2PA references reduced to 0, with the trainedAlgorithmicMedia flag removed. What Calabi does not claim to touch is Veo's invisible pixel watermark — that layer is outside metadata.

The steps:

Upload your Veo export.
Calabi automatically strips the metadata AI signals and injects a phone-capture identity.
Review the forensic proof card.
Download the cleaned file.

FAQ

Does Calabi remove Veo's invisible pixel watermark?

No. Calabi cleans the metadata and provenance layer — the C2PA manifest, XMP AI flags, and encoder fingerprints. Veo's invisible per-pixel watermark is a separate layer; a re-encode may disrupt some patterns, but no metadata tool reliably removes it, and results vary.

Is removing the C2PA manifest worth it if the pixel watermark stays?

The C2PA manifest and XMP flags are the signals many platforms scan first, so clearing them removes the most explicit "made by AI" record from the file. It's an honest partial step, not a guarantee — Calabi shows you exactly what was stripped on the proof card.

Can Calabi guarantee my Veo clip won't be detected?

No. Results vary by platform and source model, and Veo's pixel watermark is designed to persist. Calabi removes the metadata and provenance layer, but no tool can promise a platform outcome.

Try Calabi free at calabilabs.com — 10 cleans, no card.