Why Did Compressing My Image Make It Bigger?
It happens more often than people expect, and it is not a bug in the tool. Here is what is really going on — and what a compressor should do about it.
You compressed a photo, and the result is larger than the file you started with. That feels like the tool is broken. It is not, and understanding why tells you a lot about which tool to trust.
Reason 1: the file was already compressed
JPEG quality settings are not absolute. A file saved at "quality 70" by one app may look and measure differently from "quality 70" in another. So when you take an image that is already compressed at a low setting and re-encode it at a higher setting, you are asking the encoder to reproduce detail that is no longer there.
It cannot. What it can do is spend bits describing the compression artifacts of the previous pass — carefully encoding the blocky edges and ringing as if they were real image content. The file grows, and it does not look any better.
This is the most common cause, and it is why raising the quality slider on a small file usually makes things worse.
Reason 2: chroma subsampling was reduced
JPEG stores colour at lower resolution than brightness. A file might carry full colour resolution (4:4:4) or half (4:2:2), or a quarter (4:2:0). If your source is 4:4:4 and the encoder decides to write 4:2:0, that should make the file smaller — so this is rarely the cause of growth. The reverse case matters more: if a tool refuses to preserve your colour sampling and instead spends bits compensating, you can see a larger file with no real gain.
Reason 3: the original was optimised, and the new one was not
There is more to a JPEG than its pixels. Two encoders can produce identical-looking images at very different sizes depending on:
- Huffman table optimisation. Optimised tables typically save a few percent.
- Progressive vs baseline. Progressive encoding usually saves 2–8%.
- Metadata. EXIF, ICC profiles and embedded thumbnails can be tens of kilobytes.
A well-optimised q85 file can easily be smaller than a carelessly-encoded q80 file.
What a compressor should do about it
The honest behaviour is simple: if re-encoding would produce a larger file, keep the original and say so.
That is what this tool does. Before writing anything, it compares the result against your input. If the result is not smaller, you get your original file back unchanged, with a note explaining why. You should never end up with a worse file that is also heavier.
The special case: already-compressed files
For files that are already compressed, re-encoding at a similar quality achieves almost nothing. But there is one thing that genuinely helps, and it costs no quality at all: rewriting only the container.
The metadata block and the entropy coding live in separate parts of the file from the compressed image data. They can be rewritten without touching a single pixel:
- Strip metadata (EXIF, GPS, ICC) — pure size reduction, image data bit-for-bit identical.
- Re-encode the entropy layer with optimised Huffman tables and a progressive scan — also bit-for-bit identical pixels.
Measured on typical camera-style JPEGs, that combination removes about 4% and every decoded pixel stays exactly the same. Small, but real, and free.
The practical rule
If you still have the original export, go back to that and compress once from there. Repeatedly compressing an already-compressed file is the one workflow that reliably makes things worse — larger files, more artifacts, and no way to undo it.