Why Adobe Stock rejects photos for quality — and how to catch it first

You wait three days. The email says "quality issues." It doesn't say which file, which problem, or what to do differently. Here's what that phrase actually covers, and how to find those problems on your own machine before you submit.

Adobe's rejection emails are deliberately generic. One reason is scale — reviewers process enormous volumes and can't write personal feedback. Another is that a single batch may contain several different problems, and the system collapses them into one category. Either way, you're left guessing.

After a few hundred submissions, the guessing stops being necessary. Quality rejections cluster into a small number of measurable problems, and almost all of them can be detected before upload — not with taste or opinion, but with arithmetic on the pixels.

What "quality issues" actually means

1. Focus and sharpness

The single most common quality rejection. Reviewers zoom to 100%, and what looks acceptable on a laptop screen often falls apart at full resolution. This includes front- or back-focus (the subject isn't where the focal plane is), motion blur from a shutter speed too slow for the subject, and general softness from a lens used wide open or a photo pushed too hard in post.

Sharpness is measurable. The standard approach is the variance of the Laplacian — run an edge-detection kernel over the image and measure how much variation it produces. Sharp images have high variance, soft ones low. A more reliable version compares the image against a deliberately blurred copy of itself: if blurring changes almost nothing, the image was already blurry.

2. Noise and over-processing

Noise gets rejected, but so does aggressive noise reduction — smeared, waxy detail where texture used to be is just as much a rejection as grain. High-ISO night work is the usual culprit on both sides: too much grain, or too much cleanup.

The practical rule: do noise reduction manually and conservatively, treating luminance and colour noise separately. Colour noise can be removed aggressively without much cost. Luminance noise is where detail lives — take too much and the reviewer sees plastic.

3. Exposure problems

Blown highlights with no recoverable detail, blocked shadows, or a flat image lacking any real black or white point. This is straightforward to check numerically: look at the brightness histogram and see how much of the frame is clipped at either end, and whether the tonal range spans a reasonable spread.

4. Camera shake and stabilisation issues (video)

For clips, the equivalent of focus is stability. Handheld footage with micro-jitter, rolling-shutter wobble on drone shots, or jello from vibration will be rejected even when every individual frame looks sharp. This is measurable too: compare consecutive frames and calculate the shift between them. Consistent small displacements mean shake.

5. Resolution and technical minimums

Adobe requires at least 4 megapixels for photos and Full HD for video. Cropping an already-small file to fix composition can push you under the line without noticing — worth checking before rather than after.

6. Duplicates and near-identical series

Not strictly a quality issue, but it produces rejections that feel like one. Submitting eight frames of the same sunset thirty seconds apart gets flagged as duplicate content. Getty's own contributor guidance says this plainly: choose the strongest file from a series and submit that one.

A real example

This is a real file from my own portfolio. Shot at golden hour on the Aegean coast, handheld at water level, at a shutter speed that should have been fast enough. It looked good on the back of the camera and it looked good in Lightroom at fit-to-screen.

At 100%, the foreground surf is soft — the focal plane landed slightly behind it, and the wide-open aperture left no margin. Measured, its sharpness score sits well below the range where my accepted files land. That's information I had before I uploaded; I just wasn't looking at it.

How to check before you submit

You can do most of this manually. Zoom every file to 100% and inspect focus. Check histograms for clipping. Read shutter speeds against your subject and focal length. Compare near-identical frames side by side and pick one. Verify pixel dimensions after cropping.

That's roughly two minutes per file if you're disciplined about it. On a 50-file shoot, fine. On the 825 files I came back with from one week of shooting, that's most of a working day — and by file three hundred your eyes stop being reliable anyway.

The alternative is to let software do the measuring and only look at what it flags. That's what I built StockMeta for: it scores every file locally — sharpness, exposure, noise, resolution, shutter speed, video shake — and sorts them into keep review reject with the specific reason listed. Near-identical frames get grouped so you only consider the best one from each series.

No AI is involved in that part, deliberately. It's computer vision and arithmetic running on your own machine — nothing is uploaded, and no model is deciding whether your photo is any good. Composition, timing, and whether the shot is worth selling remain entirely your call. The software only answers the question a reviewer will ask at 100%: is this technically sound?

What it doesn't solve

Rejections aren't all technical. Intellectual property refusals — visible logos, recognisable trademarks, identifiable private property — are a separate category, and no local analysis will catch a brand name on a boat hull. Commercial viability is another: a perfectly sharp photo of a saturated subject may be accepted and never sell. And editorial submissions have their own caption and content rules entirely.

But the technical bucket is the biggest one, and it's the one you can close before it costs you three days.