Does Imgur Compress Images? It Made Mine Bigger

Yes and no, and the result surprised me. Does Imgur compress images? It re-encodes them, but on 2026-07-18 my 4.13 MB iPhone photo came back at 4.80 MB, 16.1% larger than it went in, with the EXIF metadata stripped. The transparent PNG I sent through alongside it came back 18.8% smaller. Imgur processes your file. It does not reliably shrink it.

Scope first, because it matters here more than usual: that is two reference files, one day, one direct upload to imgur.com, one download of the file Imgur served back. It does not prove what Imgur does to every image, and Imgur's own rules change with your login state and your file type. Everything below says which part is measured and which part is documented.

Does Imgur Compress Images? What Two Reference Files Actually Did

It re-encodes them, and re-encoding is not the same thing as compressing. One file got smaller, one got noticeably bigger, and both came back at their original pixel dimensions. That split is the whole story on this page, and no other result in this test series looks like it except Slack.

The two reference files

Every platform in this series meets the same two files, so the numbers line up across pages. REF-A is a real iPhone photo: 4,332,705 bytes, 4032x3024, JPEG, with a full EXIF block. REF-B is a photographic cutout with real transparency: 1,863,465 bytes, 1240x930, PNG with an RGBA alpha channel. Using one pair of files across a dozen apps is the only honest way to say Imgur is rougher on a photo than Discord is, because the input is byte for byte identical in both tests.

Bigger is still a re-encode

A file that comes back 670 KB heavier has obviously been through something. Two facts prove it was not just passed along: the byte count changed, and the EXIF block that my camera wrote into the original was gone from the file Imgur served. A host that stores your bytes and hands them back cannot strip metadata, because it never opened the file. Imgur opened mine, decoded it, and wrote a new JPEG.

That new JPEG was written at a quality setting higher than what my camera had already used. When you re-encode an already lossy JPEG at a higher quality target, the encoder spends more bytes describing the compression artifacts that are now part of the image. You pay for detail that was thrown away before you ever uploaded. The result is a bigger file that contains slightly less information than the one you started with, which is the worst trade in image handling.

What this result does not prove

The measurement store does not record whether that upload ran anonymously or through a logged in account, and Imgur's documented thresholds differ between the two. It also covers the direct web upload path only, not the API, not a comment attachment, and not a third party client like ShareX. Every app in this series gets the same treatment on the hub for which apps compress your images, including the parts each test leaves out.

Why Did My Photo Come Back 16% Bigger?

Because a re-encode at the wrong quality target inflates a JPEG instead of shrinking it. This is not an Imgur bug and it is not rare. Slack did the same thing to the same file, returning it 10.9% larger, and the mechanism is identical in both cases.

Generation loss, in bytes

JPEG is lossy, meaning the encoder permanently discards image information to save space, and the discarded data never comes back (MDN's definition of lossy compression is the clean version of this). A camera JPEG has already made that trade once. When a platform decodes it and encodes it again, two things happen at the same time. Quality goes down, because a second lossy pass discards more. And size can go either way, because the second encoder is not aiming at the size of your file, it is aiming at its own quality setting.

If the platform's setting is more aggressive than your camera's, you get a smaller file and visible softening. If it is less aggressive, you get a larger file and slightly worse detail, which is what happened to mine. The full mechanics of that tradeoff are in the guide on lossy versus lossless compression.

Why the PNG went the other way

REF-B dropped 18.8%, from 1.78 MB to 1.44 MB, and stayed a PNG at 1240x930. PNG compression is lossless, so a re-encode does not discard pixels, it just repacks them. Imgur's encoder packed that file better than the tool that produced it did, and the image on the other end decodes to the same pixels. A PNG getting smaller on upload is normal and costs you nothing.

The one thing my test did not separately check on that file was the alpha channel, so I am not going to claim your transparency is safe on Imgur. If transparency is the point of your file, compressing a PNG without losing transparency covers how to verify it yourself before and after.

What Are Imgur's Upload Limits and Compression Thresholds?

Imgur's rules are threshold based, not universal, and the imgur image size limit you hit depends on whether you are logged in. These are Imgur's own published numbers, not mine.

Those thresholds and conversions come from Imgur's own help article on file types and size limits, and the same table is quoted verbatim in a ShareX issue thread on Imgur upload limits if you want a second copy of it. The 50 per hour cap comes from Imgur's FAQ for forum users.

My file landed between the two thresholds

Here is where the documentation and the measurement have an argument worth reporting honestly. My 4.13 MB photo sits above the 1 MB anonymous threshold and below the 5 MB account threshold. So one of two things was true. If that upload was anonymous, Imgur's documented lossy compression applied to a 4.13 MB file and produced a 4.80 MB file, which means the rule fired and still did not make anything smaller. If it was logged in, the file was under the threshold and should not have been touched at all, yet the EXIF was stripped and the bytes changed anyway.

Either branch tells you the same practical thing: the published thresholds describe when Imgur decides to process your image, not whether the output will be smaller than the input. Nobody in the search results for this question makes that distinction, because nobody else measured it.

The imgur upload limit that actually bites

For most people the binding constraint is not the 20 MB ceiling, it is the PNG conversion at 5 MB. A large screenshot, a UI capture, or an exported diagram can cross 5 MB without feeling large, and above that line Imgur turns your lossless PNG into a lossy JPEG. Hard edges and small text are exactly what JPEG handles worst, so that conversion is where visible damage usually shows up.

Does Imgur Compress Photos and Screenshots the Same Way?

No, and my two files prove it. Does Imgur compress photos the way it compresses graphics? The photo came back 16.1% larger with its metadata gone, and the PNG came back 18.8% smaller with its dimensions intact. Same host, same day, opposite outcomes.

Photographs: the risky case

A camera photo is already a lossy JPEG, so anything Imgur does to it is a second lossy pass. You lose a little detail every time, and as my test showed, you are not even guaranteed to save space in exchange. The safe move with a photo is to decide its size yourself before you upload, so Imgur has nothing to gain by re-encoding it. Run the photo through the compressor first and a typical phone photo comes back 60 to 80% smaller with nothing visible lost.

Screenshots and graphics: watch the 5 MB line

PNG is the right format for screenshots, UI captures, and anything with hard edges or flat color, and Imgur re-compressed my PNG losslessly without touching its dimensions. That is good behavior. It stays good behavior right up until the file crosses 5 MB, at which point Imgur converts it to JPEG and the advantage you chose PNG for is gone. Keep graphics under 5 MB and the format you uploaded is the format that gets served.

Animated files are a different product

GIFs and video over 2 MB become GIFV, which is Imgur's own delivery format rather than a recompressed GIF. That is a format change with different playback behavior, not a quality setting, and it is the one conversion in the list that has nothing to do with image compression at all.

What Imgur Changes: File Size, Dimensions, Format, and EXIF

Four properties, four different answers, and only one of them is what people expect.

Dimensions survive, which is genuinely good

Imgur kept all 12 megapixels of my photo and every pixel of the PNG. That puts it well ahead of the platforms that quietly resize: LinkedIn cut the same photo to 480x360, and Facebook took it to 2048x1536. If you need the full resolution to still be there when someone downloads your link, Imgur delivers that.

EXIF is stripped, which is mostly good

The metadata block your camera writes carries GPS coordinates, the camera model, and the exact timestamp of the shot. Imgur removed it from my photo. For a public image host that is the right default and worth knowing, because anything you post to Imgur is one right click from a stranger's hard drive. It also means Imgur is not the place to check whether your metadata survived a workflow, because it will not have.

Does Imgur reduce image quality on a photo you upload twice?

Yes, and it compounds. Does Imgur reduce image quality on a single pass? A little, invisibly at normal viewing size. But every round trip through a lossy re-encode costs you again, so downloading an Imgur copy and re-uploading it is how images turn to mush over a few generations. Keep your original somewhere Imgur is not, and upload from that original every time.

The Stack Exchange Imgur Instance Plays by Different Rules

If you arrived here from a Stack Exchange site, your numbers will not match mine, because Stack Exchange runs its own Imgur backed image pipeline with its own thresholds. This is the single most confusing thing about the topic and almost nobody separates the two.

Users on Meta Stack Exchange documented a resolution halving that triggers around 760 KiB on the Stack Exchange instance, well below anything in Imgur's public documentation, in a long running Meta Stack Exchange thread about SE's Imgur instance reducing resolution. That thread has never been given an official answer.

The practical rule is simple. Imgur's published limits and my measurement describe imgur.com. Images you drag into a Stack Exchange answer box go somewhere else that happens to be built on Imgur, and their aggressive downscaling is Stack Exchange's behavior, not the one you get from a direct upload.

How to Upload to Imgur Without Losing Quality

Four steps, in the order that matters. This is how to upload to Imgur without losing quality in practice, given that Imgur will process the file no matter what you do.

1. Log in before you upload

A logged in upload raises the documented lossy compression threshold from 1 MB to 5 MB. That is a five times larger safe zone for one click, and it is the cheapest quality lever available. It costs nothing and it is the only Imgur side control you actually have.

2. Decide the file size yourself, first

The reliable version of this is to compress the image on purpose before it goes up, so it arrives well under the relevant threshold and Imgur has no reason to intervene. Choosing quality 80 yourself is a decision you can see the result of. Letting a platform re-encode at a setting you cannot inspect, and possibly hand you a bigger file, is not.

3. Keep PNGs under 5 MB and photos as JPEG

Do not upload a photograph as a PNG hoping to dodge the lossy pass, because a PNG of a photo is enormous and will sail past 5 MB, where Imgur converts it to JPEG regardless. Send photos as JPEG and graphics as PNG under 5 MB, which is the format split that matches how each encoder was designed anyway.

4. Check the served file, not the preview

Open the direct image link and download it, then compare the byte size and dimensions with your original. The Imgur page view is a display copy, and judging quality by looking at it in a browser tab is how people conclude the wrong thing in both directions. Two minutes with the actual downloaded file settles it.

How Imgur Compares to the Other Apps I Measured

Imgur sits at the wrong end of this chart, and the chart is the point: the same 4.13 MB photo, the same day, eight different platforms, measured on disk after each round trip.

Two platforms in this whole series handed the photo back bigger than they got it, and both are places people paste screenshots all day. Slack did it by 10.9% and Imgur did it by 16.1%. Meanwhile Discord returned the same file within 0.3%, and Facebook stripped 86.3% out of it because a social feed has no use for 12 megapixels. The split is not gentle versus rough hosts. It is whether the platform serves your file to you or publishes it to a feed, and Imgur is trying to do both at once.

Make the Size Decision Before Imgur Does

Imgur is one of the few places a file can come back heavier than it went in, and there is no setting anywhere in the product that stops it. What you do control is what you hand it.

Drop the photo on the compressor first, at quality 80, and a typical phone photo comes back 60 to 80% smaller with nothing visible lost, which puts it comfortably under every threshold in Imgur's table. It strips the EXIF block before anyone else can read it, and handles thirty images at a time up to 25 MB each, free. The full method is in the main guide on how to compress images.