Remove Content Credentials From an Image
"Content Credentials" is Adobe and the C2PA alliance's name for the provenance information that can be embedded inside an image — who made it, what tools were used, whether AI was involved, and how it was edited. Underneath, Content Credentials are stored using the C2PA technical standard, typically as a JUMBF box embedded inside the image file.
If you want to remove Content Credentials from an image — for example, to minimize personal metadata before posting publicly — the simplest approach is to re-encode the image. The noc2pa.com tool does exactly that: drop your PNG or WebP into the converter, and it produces a JPG with embedded provenance metadata stripped during the re-encode. Everything happens in your browser; the image is not uploaded anywhere.
Before you remove provenance: Content Credentials exist for a reason — they help viewers and platforms understand how an image was created. Don't strip them in order to deceive an audience that has a reasonable expectation of provenance. See our Acceptable Use Policy for the line we draw.
Why a re-encode removes Content Credentials
Content Credentials live inside a structured metadata block bolted onto the image file. JPEG stores it inside an APP11 marker; PNG uses a custom chunk; WebP uses RIFF chunks. When you re-encode an image — whether by saving from a graphics program, by piping through a converter, or by drawing the pixels onto a canvas and exporting fresh JPG bytes — only the pixel data is preserved. The provenance container is not carried forward, so the new file does not contain the original C2PA manifest.
That's a side effect of re-encoding, not a sophisticated bypass. The same thing happens whenever an image is screenshotted, cropped in many apps, or compressed by a messaging platform that doesn't preserve metadata.
How to remove Content Credentials with noc2pa
- Open noc2pa.com.
- Click Choose Image and select a PNG or WebP file.
- The image is decoded onto a canvas and re-encoded as JPG, locally in your browser.
- Click Download as JPG. The output JPG does not carry the original Content Credentials manifest.
- To confirm, run the output through the C2PA Checker or upload it to the official Content Credentials Verify tool.
What this won't do
- It won't reverse public, distributed provenance. If a Content Credentials manifest references an externally-stored claim (a "soft binding" or cloud-side record), removing the embedded manifest from your local file does not delete that external record.
- It won't fool every detection method. Some platforms can identify AI-generated images via pixel-level signals (invisible watermarks, statistical fingerprints) independent of any embedded metadata.
- It won't make a re-encoded JPG identical to the original. JPG is lossy. If you need pixel-perfect output, this isn't the right tool.
Removing Content Credentials with other tools
If you're working on the command line, re-encoding with ffmpeg, cwebp, or ImageMagick typically drops C2PA data the same way a browser canvas re-encode does. ExifTool can remove many metadata fields directly without re-encoding, but its handling of C2PA is format-specific — see Can ExifTool remove C2PA metadata? for the details.
Frequently asked
- Is the file uploaded to your server?
- No. Conversion happens in your browser using
FileReaderand<canvas>. The bytes never leave your device. - Does this work on AI-generated images?
- If an AI-generated image carries Content Credentials in a supported PNG or WebP file, the re-encode will drop the embedded manifest in the output JPG. Don't use this to misrepresent how an image was created.
- Why JPG output specifically?
- JPG is the most broadly compatible format for sharing photographs, and the pure-pixel re-encode is straightforward in a browser. Other re-encode targets would work too; we picked JPG for ubiquity.
- Will the new JPG ever be detected as having had Content Credentials before?
- The output JPG itself doesn't contain the original manifest. Whether another party can infer that the image once had Content Credentials depends on signals outside the file (pixel-level watermarks, hash matching against a published claim, etc.). This tool addresses the embedded metadata only.