KNOW THE DIFFERENCE
SynthID vs C2PA: Watermarks and Metadata
A Remove AI Label guide to distinguishing removable file attachments from signals inside image content.
Two different locations
EXIF, XMP and embedded C2PA records occupy file containers. SynthID is embedded within media content. Deleting an attachment does not remove a pixel watermark or establish that an image is camera-original.
Choose the right check
Remove AI Label inspects supported metadata presence and readable fields. A credential verifier checks signatures; a suitable watermark detector checks its specific signal. None is interchangeable with a platform’s classification or a visually unchanged preview.
Verify the actual output
Try the local PNG sample, inspect its parameters field, then compare the cleaned copy. This demonstrates supported metadata removal only. Encoded pixels, color profiles and necessary orientation remain intact. Cleanup removes supported descriptive containers together, not one selected field.
Keep claims precise
Say which metadata was checked and removed. This local tool cannot guarantee platform AI-label changes or remove SynthID. JPG, PNG and WebP are supported; HEIC, AVIF, GIF, PDF, video and audio are not. Keep originals when provenance matters.
Separate evidence from a progress screen
A generated PNG sample contains a known parameters field. Its original report can show that value and its cleaned report can show the field gone. This is reproducible evidence about a supported file attachment. A progress animation, generic preview or sign-up screen alone does not establish that a pixel signal was measured or removed. Verify the actual downloaded output with appropriate tools. Preserve original files, and keep any disclosures required by the place where you publish.
Further reading: C2PA specification · Google SynthID