Research

Counted data on AI product photography

Every report here counts a dataset Berrys publishes, using one calculation that the page, its markdown version and its FAQ all read. Denominators are stated, assets are counted as present only when a path exists, and anything built on fewer than 20 records is dropped rather than rounded up.

Reports are recomputed when the underlying dataset changes · Dataset licenses and usage: /ai.txt

Published reports

Dataset audit · Updated 6 November 2026

AI Model Pose Library Statistics 2026: 100 Poses, 500 Angles

A full audit of the open pose library: 100 poses as 500 camera-angle records, 500 ControlNet depth maps at 100% coverage, 426 written pose briefs (85.2%), and the 7 posture families the dataset actually contains.

Read the report →
Nothing on this index is an estimate. If a number is not derivable from a published Berrys dataset, it does not appear in a report — which is why the index is short and specific rather than long and general.

How these reports are built

One calculation per datasetDenominators published20-record minimum sampleAssets counted, never inferredExclusions listed

A report starts from a dataset Berrys owns and publishes — the pose catalog at /poses-full.json, the marketplace spec tables behind the compliance checker, and the node workflows documented in the journal. The calculation lives in a pure module, so the HTML page, the .md version, the FAQ and the machine index all read the same numbers and cannot drift apart.

Want the dataset behind a report?

The pose catalog is public JSON today, and the workflows that use it run on the Berrys canvas. Start with the free tier, or brief Studio to run the production for you.

Chat with us