auto_awesome Features

How PixPick culls your photos

PixPick is desktop software that does the first pass of culling for you. Point it at a folder of JPG, HEIC or RAW files and it flags blurry shots, closed eyes, bad exposure and near-duplicates, scores every frame and gives it a 1–5 star rating. You review, adjust, and export. Everything runs on your own computer: photos are never uploaded.

The workflow

  1. 01

    Pick a folder

    Choose the shoot folder. Sub-folders, RAW and HEIC are supported.

  2. 02

    AI analyses

    Every photo is checked and scored locally, a few hundred milliseconds each on an ordinary CPU.

  3. 03

    Review

    Filter by star rating or problem type. Fix any rating with the 1–5 keys, like in Lightroom.

  4. 04

    Export

    Write star ratings to XMP, or sort keepers into folders with hard links or copies.

What it checks in every photo

Photos that hit any of these problems drop straight to 1 star, so you can review or discard them in one go.

motion_photos_off

Blur and missed focus

Sharpness is measured with two methods (Laplacian variance and wavelet analysis), so both camera shake and soft focus are caught.

visibility_off

Closed eyes

Faces are located and mapped with 468 landmarks; eye openness is measured for each face. Tiny background faces are ignored.

center_focus_strong

Focus on the eyes

For portraits, the sharpness of the eye region is checked separately, because a sharp ear with soft eyes is still a miss.

palette

Over- and under-exposure

Clipped highlights and crushed shadows are measured from the histogram.

grid_view

Near-duplicates and bursts

Similar frames are grouped by perceptual hash. Within each group the best-scoring shots are kept and the rest marked as duplicates.

error

Empty frames

Shots with no clear subject (accidental shutter presses, lens-cap shots) are flagged.

How the score and stars are worked out

Technical quality

Sharpness, exposure and colour saturation.

Composition

Rule-of-thirds placement, horizon tilt and subjects cut off at the frame edge.

Aesthetics

An open-source CLIP vision model with the LAION aesthetic predictor rates how appealing the image looks.

These combine into a final score, which maps to stars: 85% and up is 5 stars, 75% is 4, 60% is 3, 40% is 2, below that 1. A Strict mode raises every threshold when you need to be tougher, for example on a paid wedding delivery.

Describe the shot you want

Type a description such as "black and white portrait with strong light and shadow" and photos that match it are pushed up the ranking. The same CLIP model reads both your words and the images, on your computer.

Review like in Lightroom

  • ← / → to move, 1–5 to rate, 0 to go back to the AI rating
  • P to pick, X to reject, F to favourite, Space for 1:1 zoom
  • Ctrl+Z / Ctrl+Y to undo and redo the last 50 actions
  • Filter by star rating or by problem: blurred, closed eyes, duplicates…
  • Rejected photos go to a recycle bin; your original files are never touched

Export to your editor

  • Star ratings written to XMP sidecar files, which Lightroom Classic, Capture One and Bridge can read
  • Sort keepers into folders by star rating or category
  • Hard links on the same drive (instant, no extra disk space) or copies

Specs

Platforms
Windows 10 / 11 (64-bit), macOS
File formats
JPG, PNG, TIFF, WebP, HEIC, RAW (CR2, CR3, NEF, ARW, RAF, DNG, ORF, RW2)
Where it runs
On your computer. No GPU required. Photos, file names and EXIF are never uploaded.
Internet
Only to sign in and check your trial or membership.
AI models
YOLOX, YuNet, MediaPipe Face Mesh, OpenCLIP ViT-B/16, LAION aesthetic predictor (all open source)
Price
Free 3-day trial with every feature, then a monthly membership

Try it on your last shoot

Every feature is free for 3 days after you sign up.