Free image utility

Compare image metadata before and after export.

Did your saved prompt, seed or ComfyUI workflow survive the export? Upload the original and the exported copy and see exactly which creation details our reader finds in each, side by side.

Two still images, up to 10 MiB each. Files are read on our server to return this result. No AI model call, account or saved public report. Privacy

What this comparison reads

Supported records include AUTOMATIC1111 PNG parameters and compatible EXIF UserComment data, supported ComfyUI PNG execution graphs and InvokeAI PNG generation records. We compare the prompt, negative prompt, recorded model and available settings, and ComfyUI editor and API graphs each get their own fingerprint.

Image dimensions and saved generation dimensions are tracked separately, so a resize that changes the picture while leaving the recorded generation size untouched shows up exactly that way. The tool reads records as saved; it reconstructs no missing instructions and makes no claim about which model created the picture.

EXIF camera fields, GPS, complete XMP records and Content Credentials signatures each have their own dedicated tool. For signed records use the C2PA checker. To copy a saved prompt use the image prompt extractor.

Our export experiment: 18 tests, downloadable files

We attached three synthetic record types to a code-drawn image and ran six explicit operations on each. The three byte-for-byte copies kept every supported creation field; every other operation changed something. 15 of 18 outputs had exactly the same decoded RGBA pixels as their original, proving that matching pixels never guarantee matching records.

These are measured results for Sharp 0.35.4, and we say exactly that: they describe the record formats we constructed, not the AUTOMATIC1111, ComfyUI or InvokeAI applications themselves. The prompts and model names are demonstration data, not claims about the image’s origin.

Observed 2026-09-19 · 18 local transformations · 3 retained every readable creation field
Input recordOperationFields retainedSame pixels?Files
AUTOMATIC1111 parametersCopy original bytes8 / 8Yes, exact RGBA matchOriginal · Export
AUTOMATIC1111 parametersPNG re-encode (default metadata)0 / 8Yes, exact RGBA matchOriginal · Export
AUTOMATIC1111 parametersPNG with keepMetadata0 / 8Yes, exact RGBA matchOriginal · Export
AUTOMATIC1111 parametersJPEG export, quality 900 / 8NoOriginal · Export
AUTOMATIC1111 parametersLossless WebP export0 / 8Yes, exact RGBA matchOriginal · Export
AUTOMATIC1111 parametersRaw pixels to a new PNG0 / 8Yes, exact RGBA matchOriginal · Export
ComfyUI editor and API graphsCopy original bytes2 / 2Yes, exact RGBA matchOriginal · Export
ComfyUI editor and API graphsPNG re-encode (default metadata)0 / 2Yes, exact RGBA matchOriginal · Export
ComfyUI editor and API graphsPNG with keepMetadata0 / 2Yes, exact RGBA matchOriginal · Export
ComfyUI editor and API graphsJPEG export, quality 900 / 2NoOriginal · Export
ComfyUI editor and API graphsLossless WebP export0 / 2Yes, exact RGBA matchOriginal · Export
ComfyUI editor and API graphsRaw pixels to a new PNG0 / 2Yes, exact RGBA matchOriginal · Export
InvokeAI generation recordCopy original bytes10 / 10Yes, exact RGBA matchOriginal · Export
InvokeAI generation recordPNG re-encode (default metadata)0 / 10Yes, exact RGBA matchOriginal · Export
InvokeAI generation recordPNG with keepMetadata0 / 10Yes, exact RGBA matchOriginal · Export
InvokeAI generation recordJPEG export, quality 900 / 10NoOriginal · Export
InvokeAI generation recordLossless WebP export0 / 10Yes, exact RGBA matchOriginal · Export
InvokeAI generation recordRaw pixels to a new PNG0 / 10Yes, exact RGBA matchOriginal · Export

Download results CSV · Full observations and SHA-256 hashes · Example license and scope

Repeat the experiment

The downloadable observations record the exact Node, Sharp and libvips versions and the call used for each operation, so you can reproduce every row. For a compact standalone exercise, compare a byte copy with a PNG re-export using the public reproduction script and files.

npx tsx scripts/create-metadata-retention.ts

Start with each original PNG. Copy its bytes, re-encode to PNG, re-encode with keepMetadata(), export JPEG at quality 90, export lossless WebP, and rebuild a PNG from decoded pixels. Then run the same reader on both files and compare their decoded RGBA buffers. Exact calls, output files and field-level results are in the JSON download.

Sharp documents that its default output strips metadata. In our fixture run, keepMetadata() also left these custom PNG creation records behind. That is a measured result for this version and these records, stated precisely rather than stretched into a claim about every metadata field. Sharp output documentation.

What this experiment proves, and what it leaves open

We measured six local operations and report exactly those. We did not upload files to social networks or test browser downloads or operating-system screenshots. Rebuilding a PNG from raw pixels is an explicit pixel-only operation, not a measured screenshot workflow. C2PA signature survival, detector accuracy and generator authorship were outside this run.

Keep the original file whenever you need its saved creation details. Test your own export path with both files above: a format name alone never tells you which fields survived, and this tool shows you precisely.