An image can lose metadata and still grow if the tool also changes how the picture is encoded. The final size combines image data, attached records and file structure. Savings in one part can be outweighed by growth in another.
A larger output is worth investigating, but it does not automatically mean metadata removal failed. Check the records and the encoding separately.
Think of the file as two ledgers
For a simplified comparison, separate the bytes that represent the picture from the optional records around it. If a metadata-only editor removes a small record while retaining the picture’s encoding, the savings may be modest. If an exporter rebuilds the picture, the outcome can be very different.
Illustrative accounting. Numbers below are hypothetical, not Deprint measurements.
Before
What this means
The metadata is only a small part of this example’s size.
Metadata-only edit
What this means
Removing 20 KB cannot explain a dramatic change in image encoding.
Re-encoded export
What this means
A different encoder or setting can outweigh the metadata savings.
The numbers in the illustration are hypothetical. Their purpose is to show the accounting: removing 20 KB cannot offset a 300 KB increase in the encoded picture.
Common reasons for the increase
| Change | Why size may grow | What to compare |
|---|---|---|
| JPEG converted to PNG | Lossless storage preserves the decoded rendition instead of using the prior lossy representation | Input and output formats |
| Different PNG encoding | Compression strategy, filtering or representation can differ | Encoder options and actual samples |
| Higher-quality lossy export | More information may be retained in the new encoding | Settings and visible quality |
| Palette changed to full color | A compact indexed representation may be expanded | Color representation |
| Added transparency or changed bit depth | More information may need representation | Alpha channel and sample depth |
| Larger dimensions | More samples are stored | Width and height |
These are possible causes, not a diagnosis of a particular tool. Inspect the actual output before choosing an explanation.
Why PNG does not mean “smallest”
PNG provides lossless compression for its supported image representation. That means preservation under the format’s rules, not minimum file size across all formats and encoders.
A detailed photograph that was efficiently stored as JPEG can become much larger when its decoded samples are saved losslessly. The PNG faithfully stores the already-decoded rendition, including artifacts introduced earlier. It does not recover the original pre-JPEG detail.
Even within PNG, two encoders can produce different byte counts while preserving equivalent image samples. The visible result can remain unchanged. File-hash inequality is therefore expected and does not alone identify a quality change.
A small diagnostic checklist
First compare format, dimensions and file size. Then inspect metadata in both files. Finally compare appearance and, when required, decoded samples with a consistent procedure.
If the output grew but the targeted GPS fields disappeared, the privacy operation may have succeeded while the encoding became less compact. If the pixels changed, investigate resizing, color conversion, orientation or lossy export. If the target fields remain, that is a separate cleanup failure.
Keep these findings in separate columns. “Bigger file” is not a complete quality assessment, and “metadata removed” is not a compression claim.
A practical example
Suppose you need to share a photograph without location tags and under an attachment limit. A tool exports a clean PNG that is larger than the source JPEG. You now have two requirements: remove the sensitive fields and meet the size limit.
A suitable metadata-aware JPEG workflow may address the first requirement without a full re-encode. If additional compression is still necessary, choose it deliberately and inspect the visual result. Do not repeatedly convert the file through several formats simply because the first output was large.
This is a decision example, not a benchmark or endorsement of a particular product. The receiving service may transform the image again, so inspect its delivered copy if that copy is what others will download.
What a useful tool report should say
A clear report separates removed records, output format, dimensions, byte count and any pixel transformation. If the tool resized or re-encoded the picture, that should not be hidden behind the phrase “metadata cleanup.”
For your own records, save the original and name the export distinctly. For a published comparison, include the tool version, settings and hashes. This lets someone reproduce the byte-count result without assuming every difference is caused by metadata.
A simple table can be enough: original, metadata-only copy, resized copy and final compressed copy. The rows reveal which step changed size and which changed appearance.
Which result should you prefer?
Choose the file that meets the actual requirement. For a private archive, preservation may matter more than small size. For a web image, delivery size and acceptable appearance may dominate. For location privacy, the relevant fields must be checked regardless of the byte count.
A clean but larger file is not automatically a bad result. An impressively small file that damages text or retains sensitive records is not automatically a good one. The compression guide helps define the preservation standard before you compare tools.