Does a low score mean the file will definitely be rejected?

No. The score reflects measurable image quality only. Agencies weigh subject, composition, similars and commercial need alongside technical quality, so a low-scoring frame with a strong subject can still be accepted, and a technically perfect frame of a dull subject can still be refused.

Are my files uploaded for analysis?

No. Decoding, sampling and scoring all happen in your browser on your own machine. You can disconnect from the network and the tool still works.

How many files can I check at once?

Large batches are fine. Analysis is spread across several worker threads and results stream in as they finish, so the page stays responsive while a few hundred files are processed.

Does it check video as well as stills?

Yes. Clips are sampled at several timestamps and scored on the same noise, focus and exposure measures, with an extra flag when a clip still carries an audio track that a marketplace would reject.

Toolbox

AI approval sorter

Point it at a folder of photos and clips. Every file is measured in your browser for sharpness, noise, exposure, contrast, banding, resolution and near-duplicates, then optionally reviewed by AI for commercial appeal, trademarks, releases and generative artifacts. Files are then moved into likely-approved/ and low-chance/ with a CSV report.

1. Source folder

Up to 2,000 JPG, PNG, WebP, MP4, MOV or WebM files, including sub-folders.

2. Destination folder

This browser cannot write to disk — you'll get a sorted ZIP instead.

3. How strict should the review be?

Strict: files scoring 72 or higher go to likely-approved/.

AI review

Strict raises every technical floor, blocks releases, upscales and over-processing, and lets only the best shot of a near-duplicate set through.

AI review uses the keys saved on your API keys page. Without a key the tool still runs on technical scoring alone.

Nothing is uploaded

Pixel analysis, duplicate detection and file moving all happen on your machine. When AI review is on, only one compressed preview per file is sent to your own provider key.

What gets caught

Blur, camera shake, noise and bad exposure
Under-resolution stills and sub-1080p clips
Near-duplicate and similar-content files
Logos, trademarks and missing releases (AI pass)
Generative-AI artifacts and policy risks (AI pass)

About the approval sorter

Rejections cost more than the sale you lost: a batch that comes back marked for noise or focus burns a review cycle and pushes the whole upload back by days. The approval sorter runs the mechanical half of a stock reviewer's checklist over your files before you submit, so the obvious failures never reach a human queue.

What it actually measures

Every file is decoded in the browser and scored on the properties reviewers reject for most often: luminance noise in flat shadow areas, edge acutance as a proxy for focus, clipped highlight and shadow percentages, effective resolution after any upscaling, and colour cast in neutrals. Video files get the same treatment on sampled frames, plus a check for whether the clip carries an audio track it should not.

Each measurement produces a number, not a verdict, and the numbers are shown next to the thumbnail so you can disagree. A deliberately grainy night shot will score badly on noise; that is information, not an instruction to delete it.

How to work through a batch

1. Add the folder
Drop in the finished exports you are about to upload. Files are read locally — nothing is sent anywhere, which matters when the batch is unreleased client work.
2. Set your threshold
The pass mark is a slider. Contributors selling to picky agencies raise it; contributors uploading editorial or documentary work lower it, because grain and motion blur are accepted there.
3. Read the three buckets
Files land in pass, borderline and likely-reject. Borderline is the useful bucket: those are the frames worth opening in an editor for two minutes rather than gambling a review cycle on.
4. Export the report
Take a CSV of every score so you can compare batches over time and see whether a lens, a camera profile or an export preset is quietly responsible for your rejections.

What it cannot judge

No local tool can predict commercial demand, and none can decide whether a recognisable face or a trademarked building needs a release. It also cannot read a reviewer's mind on similars — uploading twelve near-identical frames of the same subject is a common rejection reason and only you know which one is the keeper. Treat the score as a technical gate, and keep the editorial judgement yours.

Questions about this tool

Does a low score mean the file will definitely be rejected?: No. The score reflects measurable image quality only. Agencies weigh subject, composition, similars and commercial need alongside technical quality, so a low-scoring frame with a strong subject can still be accepted, and a technically perfect frame of a dull subject can still be refused.
Are my files uploaded for analysis?: No. Decoding, sampling and scoring all happen in your browser on your own machine. You can disconnect from the network and the tool still works.
How many files can I check at once?: Large batches are fine. Analysis is spread across several worker threads and results stream in as they finish, so the page stays responsive while a few hundred files are processed.
Does it check video as well as stills?: Yes. Clips are sampled at several timestamps and scored on the same noise, focus and exposure measures, with an extra flag when a clip still carries an audio track that a marketplace would reject.