What is C2PA / Content Credentials?
C2PA is an open technical standard from the Coalition for Content Provenance and Authenticity. Adobe, Arm, Intel, Microsoft and Truepic founded that consortium. Camera makers, AI labs and major platforms have since joined it. C2PA answers one question: where did this file come from, and what happened to it along the way?
The standard's user-facing layer is Content Credentials: cryptographically signed metadata stored inside the file. A credential can name the device or model that created the image. It also records the software that edited it and the timestamps involved, plus a hash that ties the signature to the exact pixels. When every step in a file's life signs its contribution, the result is a verifiable provenance chain. That chain breaks visibly the moment anyone edits a pixel.
For AI-generated images, Content Credentials have become the disclosure mechanism that regulators and platforms rallied behind. They let a viewer confirm an image came from a specific model without trusting whoever posted it. Signed provenance is also why some creators want the record gone. It pays to know exactly what a strip removes before you run one.
Where Content Credentials live in a file
Content Credentials live in the metadata of your file. That means JUMBF boxes in JPEG, XMP packets, or dedicated chunks in PNG and WebP. They never touch the pixels. A metadata-level strip deletes the signed provenance record and leaves the image data alone. The same pass also clears any embedded previews of the credential manifest and the certificate references. The picture keeps its dimensions and composition. PNG and WebP files keep their original quality; a JPEG is saved once more at high quality, at the same size.
The removal pass targets the whole credential stack. It clears the C2PA manifest store and its JUMBF wrapper. The XMP fields that reference the manifest go too, along with EXIF blocks that may echo provenance hints. Standard metadata removal also strips technical fields such as camera settings, timestamps and GPS coordinates. The export carries none of them.
What comes back is the same image at the same dimensions, minus the signed history. The pass removes the embedded color profile (ICC) with the rest of the standard metadata, so colors follow standard display defaults. Platforms and validators that look for Content Credentials find none.
Read Content Credentials before you strip them
Reading Content Credentials and removing them are two different operations. This page keeps them separate. The inspection pass only reports what your file carries. The strip pass deletes the record, and that step is irreversible. Once the manifest is gone, the provenance chain it documented cannot be recovered from the file.
The inspection pass works as a C2PA checker in practice. It is a forensic metadata scan that reads your file and reports the provenance structures it recognizes. Those include C2PA manifests and their signers, XMP packets, EXIF records and GPS tags. What the file holds decides whether a strip is worth running. A file exported from Photoshop with Content Credentials enabled carries a far richer embedded history. A screenshot of a post carries almost none.
Both passes run in a single execution and one upload. Inspection comes first, the strip second. Nothing is removed before you see what the file carries. The strip works at the metadata layer only. The image you download keeps its dimensions and visual quality, with the signed history gone.
How it works
How to use the C2PA remover
Step 1: Upload your image
Select or drop a JPG, PNG, WebP, BMP, HEIC or AVIF file into the panel. The upload goes straight to processing storage, and nothing is published or shared.
Step 2: Inspect and strip the provenance metadata
The forensic scan reads the file and inventories what is embedded. That includes C2PA manifests, XMP packets, EXIF records and GPS tags. The scan runs before any removal happens. The metadata pass then removes the credential stack. It clears the manifest store, its JUMBF wrapper, XMP references and standard EXIF fields. The image keeps its dimensions and framing; PNG and WebP files keep their original quality, and a JPEG is saved once more at high quality. Keep an untouched copy of the original before you strip it. The signed history cannot be reconstructed afterward.
Tip
Many files arrive with no credentials left, because a screenshot or a re-export already stripped them. The inspection tells you which case you are in.
Step 3: Download the cleaned file
Preview the result and download it. The output keeps your image’s dimensions, with no signed provenance record left inside.
Tip
If the file came from a Google model, the SynthID watermark sits in the pixels and survives this pass. Pair it with a pixel-level pass.
C2PA vs SynthID: metadata vs pixels
C2PA Content Credentials are metadata: a signed record stored beside the pixels. The record disappears the moment the metadata does, and this page handles that case.
Google DeepMind's SynthID works at the opposite extreme. The signal is woven into the pixel values during generation, invisible to the eye and detectable by a classifier. SynthID also survives metadata edits, compression, cropping and mild re-rendering. Stripping C2PA has no effect on the SynthID signal. A file can leave this page with zero credentials and still carry a live SynthID inside its pixels.
For Google tools such as Gemini image generation or Imagen, pair this page with a pixel-level pass. The SynthID Remover targets the embedded watermark directly. The SynthID Detector can first tell you whether one is present. When provenance matters, check both layers.
Limitations and honest expectations
FAQ
Frequently asked questions
Common questions about this C2PA remover, from what Content Credentials are to what a strip can and cannot reach.
Explore more tools
SynthID Remover
Strip SynthID pixel watermarks from Google-generated images, the layer a metadata pass cannot reach.
SynthID Detector
Check whether an image carries a SynthID pixel watermark before you pick a remover.
Gemini Watermark Remover
Handle Gemini and Imagen watermarks at the pixel layer and the metadata layer together.
AI Metadata Remover
Apply the same strip to images whose provenance lives in XMP and IPTC rather than C2PA.
Image Metadata Viewer
Read the manifest and every other field in the file before anything is removed.
Adobe Firefly Metadata Remover
How Firefly signs an export, and what the strip leaves in place after the manifest goes.
Metadata Remover
The wider privacy pass for EXIF, GPS and editing history, alongside the provenance strip.
Bing Image Creator Metadata Remover
Clears the C2PA and IPTC records Microsoft writes into Bing Image Creator, Designer and Copilot images. The visible logo is not covered.
LinkedIn AI Label Remover
Strips the C2PA record LinkedIn reads before it applies its AI label, plus its IPTC fields.
X AI Label Remover
Strips the C2PA, IPTC and generator fields before you post on X. X automatic labelling is undocumented, so the page makes no promise about it.
All tools
Every remover and detector on one page, grouped by what each one cleans or checks.
Strip the C2PA credentials, keep the pixels
The forensic inspection shows what is signed into the file. The strip removes it in the same run. Your image comes back at its original dimensions, with no provenance record left inside.