What Actually Happens to Your Photo After You Press Upload
You choose a file and press upload. A moment later your image appears — smaller, slightly different, sometimes rotated, occasionally noticeably worse. Quite a lot happened in that moment, and almost none of it is visible.
Step One: Your Browser Reads the File
Before anything is transmitted, the page you are on gets access to the file. Most modern upload forms do at least some work here — checking the file type, rejecting anything oversized, and generating the little thumbnail preview you see.
Many services also resize the image in your browser before sending it. This is genuinely in everyone's interest: it saves your bandwidth and their processing time. But it means the first quality decision has already been made before a single byte has left your device, using settings you were never shown.
This is one reason the same photo uploaded through a website and through a mobile app can end up different. They are separate implementations making separate choices.
Step Two: Metadata Gets Stripped
Once the file reaches the server, one of the first things that usually happens is that its metadata is removed. Photographs carry a surprising amount of it: the camera model, the exact time, lens settings, and frequently the GPS coordinates of where the picture was taken.
Stripping it is good practice. Publishing an image with its original metadata intact can reveal a home address, and there is a long history of people being located this way without realising it.
Metadata also contains the orientation flag that tells viewers which way up the photo should be. If a service removes the metadata without first applying that rotation to the actual pixels, the instruction is lost while the sideways pixels remain. This is why a photo that looked perfect on your phone can appear rotated the moment it is uploaded — and why the problem is permanent once it happens.
Step Three: It Is Decoded, Resized and Re-Encoded
The server now decodes your image back into raw pixels, resizes it, and encodes it again. This is where most of the visible quality change occurs.
It rarely produces just one output. A typical service generates a whole family of sizes from your single upload — a thumbnail for lists, a medium version for feeds, a large one for full-screen viewing, sometimes several more for different screen densities. Each is a separate encode.
The critical detail is that these are almost always generated from your uploaded file rather than from anything better. If your upload was already compressed, every derivative inherits those artifacts and adds its own on top.
| Stage | What happens | Quality cost |
|---|---|---|
| Your camera | Scene captured, saved as JPEG | First lossy encode |
| Your editing | Crop or filter, re-saved | Second, if saved as JPEG |
| Browser pre-upload | Possibly resized | Sometimes a third |
| Server processing | Decoded, resized, re-encoded | Another, per output size |
| Delivery | Possibly converted to WebP | Usually negligible |
Each row is independent. This is why an image can look fine locally and poor once published.
Step Four: Someone Else Delivers It
The file that eventually reaches a visitor is usually not served by the website itself. It comes from a content delivery network — a set of servers distributed globally so that the image travels from somewhere physically near the person requesting it.
Many CDNs perform a final transformation on the way out. They inspect what the requesting browser can handle and convert the image accordingly, commonly to WebP or AVIF, both of which are smaller than JPEG at equivalent quality.
This step is generally beneficial and usually invisible. But it means two people can request the same URL and receive genuinely different files, in different formats, at different sizes, depending on their browser and their screen.
It is also why replacing an image can appear to do nothing. The old version may be cached at the edge for a long time, and the cache does not know you changed anything.
Why the Same Photo Looks Different Everywhere
Put the whole chain together and the familiar frustrations stop being mysterious.
- It looked sharper on your phone — you were seeing the original; the site is showing a re-encoded derivative.
- It went sideways on upload — the orientation metadata was stripped without the rotation being applied first.
- The colours shifted — a colour profile was discarded during processing, so the image is now interpreted differently.
- It got blurry — it was scaled to a size the layout wanted rather than the size you supplied.
- The old version is still showing — a cache somewhere is serving it.
- It looks worse than a friend's — theirs went through fewer compression generations before upload.
What You Can Actually Control
You cannot change what a platform does with your file. You can control what you hand it, and that determines how much of the pipeline works in your favour.
- Upload at close to the display size. Sending something far larger just means the service downscales it with its own settings, adding a lossy step you could have controlled yourself.
- Upload from the least-compressed source you have. Every previous save is damage the pipeline will preserve and add to.
- Fix rotation before uploading, by physically rotating the pixels rather than relying on metadata that may be discarded.
- Choose the format to match the content — never send a text-heavy screenshot into a pipeline that will encode it as JPEG.
- Keep an untouched original. Everything you publish should be exported from it, not derived from a previous export.
The through-line is straightforward: every lossy step is permanent, and they accumulate. The pipeline will add several of its own no matter what you do. The ones worth eliminating are the ones you added before it ever got there.
Frequently Asked Questions
Why does my photo look worse on a website than on my computer?
Because the version you are looking at online is not your file. It was decoded, resized and re-encoded on the server, often several times over for different display sizes. You were comparing it against the original, which never went through any of that.
Should I upload the largest possible version of my image?
Usually not. If you upload something far larger than it will ever be displayed, the service downscales it using its own settings, which adds a lossy step you had no control over. Uploading at roughly the intended display size, or twice that for high-density screens, gives a better result and a faster upload.
Why did my photo rotate when I uploaded it?
The service almost certainly stripped the EXIF metadata for privacy without first applying the rotation that metadata described. The instruction is gone while the un-rotated pixels remain. Rotating the image yourself before uploading, so the pixels are physically correct, prevents it.
Why does my updated image not appear?
Caching. Content delivery networks and browsers both hold copies of images, and neither knows you replaced the file. Most platforms provide a way to purge the cache, and changing the filename or URL sidesteps the problem entirely by making it a new asset.
Does uploading the same photo repeatedly make it worse each time?
Uploading the same original file repeatedly is fine — each upload starts from the same source. What causes cumulative damage is downloading the processed version and re-uploading that, then repeating. Each round trip adds another compression generation, which is exactly why images forwarded repeatedly degrade so visibly.