Optimising Photos for Upload: Size, Sharpness, Format

Updated: August 2026 – The photo looked crisp on your own screen, and after uploading it looks flat, mushy or oddly desaturated. This guide explains why that happens and how you optimise photos for upload – using the three levers of colour space, export size and output sharpening, each tailored to railway photo sites, photo communities and social networks.

Why photos look worse after uploading

The quality loss happens in three clearly identifiable places: colour space, scaling and compression. If you export in a wide colour space like Adobe RGB, the browser interprets the values as sRGB and renders the colours paler. If you upload a file that’s significantly too large, the server downscales it with its own algorithm, and edge detail gets lost in the process. And when the file is then compressed again, artefacts appear on exactly the structures typical of railway subjects: overhead wires, railings, fine markings.

The good news: all three effects are avoidable, with settings you set up once and simply reuse from then on.

Colour space: the invisible classic

Colour spaces describe what colour a given combination of numbers represents. sRGB is the lowest common denominator on the web – and exactly the one browsers, messaging apps and most other apps assume when no profile is embedded.

In practice that means:

• Feel free to work in a larger working colour space during development, if your software offers it.
• But always export to sRGB and have the profile embedded.
• If in doubt, check with a strongly saturated subject: a red vehicle shows colour-space errors most clearly.

If you show your photos across several channels – say, on a photo site as well as on social networks, as described in the overview on trainspotters on social media – a fixed sRGB rule saves you a lot of troubleshooting.

Export size: better to scale it yourself

The second lever is pixel size. Every platform has a maximum display width; anything above that gets downscaled server-side. That downscaling is the point where most sharpness is lost, because you have no control over the algorithm.

The rule is therefore simple: scale it yourself to a size that suits the target platform, and then sharpen for exactly that size. How large that actually is, each site states in its own upload rules – those figures are what count, not some rule of thumb from a forum. For railway photo sites with editorial review, it’s especially worth checking their specific requirements, as the practical guide to bahnbilder.de with workflow and legal questions shows.

Export profiles compared

The table below maps the typical target categories to their requirements. Specific maximum dimensions and file size limits differ by provider and change over time – check them before your first upload, and occasionally again afterwards.

Target	Colour space	Scaling	Sharpening	What else matters
Railway photo site with review	sRGB	exact to site’s specification	moderate, not visible	clean captions, no halos, correct vehicle data
Photo community like Flickr	sRGB	long edge, original size possible	subtle, for screen	tags and albums, choose licensing deliberately
Social networks	sRGB	on the smaller side, but cleanly sharpened	somewhat stronger	factor in heavy compression, check cropping
Forum post	sRGB	small to medium edge	moderate	mind the forum’s file size limit
Own website or blog	sRGB	responsive variants	separate per size	load time, modern formats optional
Archive and backup	working colour space	none	none	keep the original file unchanged

One point runs through every row: the colour space is the same everywhere. It’s the only setting you never need to decide per platform – sRGB applies to every web target. Scaling and sharpening strength differ, however, because they depend on display size, as does how metadata is handled, because some services keep it and others strip it.

The difference between editorially reviewed sites and social networks is especially notable. Sites typically expect a technically clean, restrainedly edited file and check the accompanying details too; social networks compress heavily on top and display images in varying crops. So if you show the same photo both ways, you should create two exports rather than look for a compromise.

The last point in the table is often forgotten: an upload export is a derivative, not a backup. The original stays untouched so you can later produce a different size or a revised version. How this fits into an organised workflow is shown in the RAW workflow for railway photos.

Getting output sharpening right

Output sharpening is the step most people skip – and then wonder why the edges look soft. When downscaling, neighbouring pixels get averaged together; edges that were crisp before lose contrast in the process. Output sharpening deliberately compensates for that.

Three points determine the result:

• It comes after scaling. Sharpening calculated for the full-size image doesn’t help the downscaled version.
• It’s tailored to the output size. An image for a small thumbnail needs more output sharpening than one for full screen width.
• It shouldn’t be visible. As soon as you see bright lines along a roofline, it’s too much. How to spot such halos is covered in the article on the five sliders for editing train photos.

You can roughly work out how much data this adds up to – handy if a forum or site imposes a file size limit:

Compression: why railway subjects in particular suffer

The third source of loss is JPEG compression. Put simply, it works in small blocks and preferentially discards information the eye supposedly won’t miss. That works well for landscapes and portraits. But with railway subjects, there are exactly two structures where this process typically becomes visible:

• Thin, high-contrast lines against a bright background. Overhead wires, catenary masts, railings and aerials against the sky. With too much compression, grainy edges appear there, known as mosquito artefacts.
• Soft, large-area gradients. A morning sky changes brightness only minimally across many image pixels. Too much compression flattens that into steps – visible as banding or blotches.

Neither can be fixed afterwards, because the information is already gone. Prevention is therefore the only strategy: choose a high quality level, compress only once, and never re-edit and re-save a JPEG that’s already been exported.

Why every additional save does harm

JPEG is a lossy format. If you open an exported file, change it slightly and save it again, you compress it a second time – this time based on data that’s already damaged. After two or three rounds the difference is clearly visible, especially on exactly the edges that matter in railway subjects. The rule is therefore: every output file is generated fresh from the original, never from an earlier output file.

How many pixels does an image really need?

The most common misconception is that more pixels automatically look better. What actually matters is the size at which an image is actually displayed. A photo that’s never shown wider than one screen column on a page gains nothing from the full sensor resolution – it only lengthens load time and gives the platform more material it has to scale down. Conversely, an export that’s too small is immediately visible as soft on high-resolution displays. The sweet spot is a size close to double the display width, but no more – each site states specific figures in its upload guidance.

The workflow from finished photo to upload

Common mistakes when uploading

• Exporting in Adobe RGB. The most common reason for pale colours in the browser. The file isn’t broken, it’s just being misread.
• Uploading the full camera resolution. The server then downscales it itself – with a result you don’t control and usually don’t like.
• Compressing too heavily. Block artefacts on overhead lines and sky gradients are irreversible. Better to go one level higher and accept a slightly larger file.
• Sharpening before scaling. The sharpening gets calculated away during downscaling, and the result still looks unsettled.
• Compressing twice. Re-editing and re-saving a JPEG that’s already been exported adds up artefacts. Always start from the original file.
• Forgetting captions. On photo platforms, the description also determines findability – how this works in practice is shown in the guide to Flickr and railway photos and the system in the Flickr guide to tags and groups for railway photography.

Conclusion

The difference between a photo that convinces online and one that disappoints after uploading is almost never the camera. It comes down to three settings you get right once: sRGB as the colour space, a self-chosen target size instead of server-side downscaling, and output sharpening matched to that output size. Set up a preset per platform for this, and you won’t lose any more time – or image quality – afterwards. Which shots even make it to export is a decision you make beforehand, for instance with the selection method from the guide to train photos and railway photography.