System requirements and performance

What imagic needs, and how fast it culls

The specs below are what the current builds actually require, and the speed figures are measured, not estimated: one named PC, a fixed public set of RAW files, the method written out so you can judge it.

Measured 2 October 2026 · imagic 1.4.4

Minimum and recommended

MinimumRecommended
WindowsWindows 10, 64-bit (version 1809 or later)Windows 11
macOSmacOS 14 Sonoma or later, Apple silicon or IntelApple silicon (M1 or later) on the current macOS
ProcessorAny 64-bit x64 processor, or Apple silicon6 or more cores. Culling runs on the CPU, so more cores finish sooner.
Memory8 GB RAM16 GB RAM, or more for libraries past about 10,000 photos
GraphicsNo GPU needed. Culling, editing and export all run on the CPU; a graphics card is not required and is not used for culling.
Disk1.5 GB for the app, 20 to 40 MB of previews per 1,000 photos, and room for your exports (about 6 MB per full-size JPEG from a 24 MP camera)An SSD, with the photos you are culling on it or on a fast external SSD
InternetNeeded to activate your key and, once, to download up to 60 MB of face and pose models the first time they are used. After that imagic works offline for up to 14 days between licence checks. Your photos are never uploaded.

Intel Macs get a lighter build (111 MB instead of 431 MB) without PyTorch and ONNX Runtime, because PyTorch no longer ships for Intel macOS. Import, AI culling, editing, RAW processing and export all work; the extras that need those libraries (AI subject masking, the optional perceptual quality score, photo captions, and pose grouping in the Action subject mode) are Windows and Apple silicon only.

Memory note: in a 9,790-photo library the app used about 4.8 GB after the 1.3 thumbnail fix, which is why 16 GB is the recommendation for big catalogues and 8 GB is enough for a shoot at a time.

Measured culling speed

Seconds per 100 RAW files from 24 to 27 MP cameras, on the test PC below. Lower is faster. A 2,000-frame wedding is 20 of these.

AI analysis (the cull)
38.5 s
per 100 photos
Import (previews)
7.1 s
per 100 photos
Export (full-size JPEG)
601.6 s
per 100 photos

So on this machine a 2,000-photo shoot imports in about 2.4 minutes and is analysed and sorted into keep, review and trash in about 12.8 minutes. Exporting is the slow part: every exported photo is a full RAW render, about 6.0 seconds each, and you normally export only the keepers.

Test setPhotosImportAI analysisExport
24 to 27 MP RAW, public samples1007.1 s38.5 s601.6 s
10 MP Sony ARW, one real shoot404.9 s13.7 s212.9 s

Seconds per 100 photos, median of 3 runs per set. The PC was in everyday use while it ran (a browser, music and an Android emulator in the background), so runs varied: AI analysis of the 24 MP set ranged from 23.5 to 39.7 seconds per 100. Peak memory: about 2.2 GB while exporting the 27 MP files, under 300 MB for the analysis worker.

The test PC

ProcessorIntel Core i5-13600KF (14 cores, 20 threads)
Memory32 GB DDR4
GraphicsNVIDIA GeForce RTX 3060 Ti (not used by the cull)
DiskSamsung 990 EVO Plus NVMe SSD
SystemWindows 11 Pro (build 26200)

How it was measured

Your numbers will differ with your processor, your camera's resolution and your disk. Export time grows with resolution: on this PC a 10 MP file took 2.1 seconds and a 24 to 27 MP file 6.0 seconds, so expect more for 45 to 60 MP. We have not yet timed a Mac; Apple silicon figures will be added here, dated, once measured on real hardware.

Time it on your own shoot

The free trial covers 500 exports with everything in imagic Max, and importing and culling never count against it. No card.