JPG / PNG / lossless container cleanup

Remove AI metadata from JPG and PNG

Inspect C2PA, EXIF, XMP, IPTC and prompt-bearing text first. Then create a clean copy by removing selected container sections without decoding or recompressing the image.

Reviewed August 12, 2026 by ClearTags EditorialMethodology and corrections

CONTEXT / EVIDENCE

JPG segments and PNG chunks are separate from pixels

A JPEG container can carry EXIF and XMP in application segments, IPTC data, comments and a C2PA manifest in APP11. PNG uses named chunks such as eXIf, tEXt, zTXt, iTXt and the C2PA caBX chunk. These layers can be copied or omitted independently from the encoded image data.

ClearTags parses the container in your browser and writes a new file without the supported sections you select. It does not draw the image to a canvas, so this JPG/PNG path avoids an additional lossy encode and preserves the existing encoded pixel payload.

FORMAT / CLEANUP RESULT

What can be inside each file type?

ClearTags reads the exact publishing file first. The result depends on the container and the sections actually present—not only on the filename extension.

File typeReadable or removable fieldsClean-copy result
JPEG / JPG
  • C2PA / JUMBF
  • EXIF
  • XMP
  • IPTC
  • comments
Selected application segments are omitted while the encoded JPEG scan data stays unchanged.
PNG
  • C2PA caBX
  • eXIf
  • XMP / IPTC
  • tEXt / iTXt / zTXt
  • tIME
Selected metadata chunks are omitted while the encoded IDAT image data stays unchanged.
MP4 / MOV
  • C2PA
  • creation time
  • encoder
  • location
  • titles / comments
Supported media streams are remuxed locally without intentional metadata carryover; codec compatibility is required.

The table describes supported container cleanup, not every proprietary field or a guarantee about a platform decision. Reinspect the clean copy before publishing.

REAL FILES / JPEG + PNG

Container cleanup tested on licensed exports

Review the reproducible results

The Lab currently covers one ChatGPT PNG plus Adobe Firefly and Photoshop JPEG files. It also has smaller synthetic fixtures for the format paths. We publish the hashes so the claim can be repeated.

Test file

ChatGPT / GPT-4o · PNG

PNG export from ChatGPT / GPT-4o; exact upstream build not supplied.

Version
Not provided by upstream
SHA-256
7cd57b4a33f0…
Test file

Adobe Firefly · JPEG

Adobe Firefly 4.0 JPEG with a declared text-to-image action.

Version
4.0.0-release-firefly_v4-main_78135.80468
SHA-256
bdc5c19d4211…
Test file

Adobe Photoshop · JPEG

Adobe Photoshop 26.11 JPEG after crop and resize.

Version
26.11.0
SHA-256
17c9e5b9f40e…
Verified result

All three copies passed the production cleanup and reinspection paths. Their JPEG scan or PNG image payload hashes stayed unchanged.

ClearTags Lab documents each fixture's scope and publishes its complete hashes.

What this format cleaner separates

C2PA · EXIF · XMP · CONTAINER
JPG

Application segments

EXIF, XMP, IPTC, comments and supported APP11 C2PA data can be removed independently from the JPEG scan data.

PNG

Metadata chunks

EXIF, XMP/IPTC text, prompt-bearing text chunks, timestamps and supported caBX provenance can be selected separately.

PIXELS

Image content

Removing container metadata does not erase a watermark or signal encoded into the visible or invisible pixel data itself.

Clean a JPG or PNG locally

WORKFLOW / LOCAL COPY
  1. 01

    Choose the final image

    Use the exact JPG or PNG intended for publishing, delivery or upload.

  2. 02

    Inspect each section

    Review C2PA separately from EXIF, XMP, IPTC, comments, prompts and timestamps.

  3. 03

    Choose full or selective cleanup

    Remove every supported metadata block or keep the authorship and rights fields you still need.

  4. 04

    Verify the derivative

    Reinspect the clean copy and compare its preview before downloading it.