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

> A complete audit of the open Berrys pose library: 100 poses, 500 camera-angle records, 500 ControlNet depth maps, and 426 written pose briefs, counted from the published dataset.

Canonical: https://berrys.io/research/ai-model-pose-library
Dataset: https://berrys.io/poses-full.json
Last updated: 2026-11-06

## What this page counts

The Berrys pose library is an open catalogue of commercial photography poses published at https://berrys.io/for/poses. This study counts the published dataset itself (lib/poses_full_sfw.json, served raw at https://berrys.io/poses-full.json) rather than sampling generated output. Every figure below is produced by one function that the HTML page, this markdown file, the FAQ and llms.txt all read, so the numbers cannot drift between versions.

## Key findings

### All 500 of 500 camera angles (100%) ship a paired ControlNet depth map, a line-art mannequin render and a JPEG thumbnail, so every pose in the library can drive pose-conditioned generation without a photoshoot.

Anchor: https://berrys.io/research/ai-model-pose-library#finding-depth-coverage
Denominator: 500 depth maps / 500 angle records

### The library holds 100 distinct poses across 7 posture families. Hand Gestures is the largest at 27 poses (27% of poses, 135 of 500 angles), followed by Seated at 23 poses (23%).

Anchor: https://berrys.io/research/ai-model-pose-library#finding-pose-families
Denominator: 100 poses categorised

### Every pose resolves to exactly 5 camera angles — Front, Top-Down, Low-Angle, Side, 3/4 View — and all 100 of 100 pose records (100%) carry the complete set, which is what lets one garment be rendered as a 5-view rotation instead of a single hero frame.

Anchor: https://berrys.io/research/ai-model-pose-library#finding-angle-set
Denominator: 100 complete pose records / 100 poses

### 426 of 500 angle records (85.2%) carry a written pose brief averaging 17.5 words, while 74 (14.8%) are visual-only and must be described from the depth map before prompting.

Anchor: https://berrys.io/research/ai-model-pose-library#finding-brief-coverage
Denominator: 426 briefed / 500 angle records

### The 426 written briefs collapse to 98 distinct descriptions, because the same posture text is reused across that pose's five camera angles — one description per pose, not one per view.

Anchor: https://berrys.io/research/ai-model-pose-library#finding-brief-reuse
Denominator: 98 distinct texts / 100 poses

### 73 of 100 poses (73%) belong to the Commercial Catalogue Series series with full-body framing, and 27 (27%) to Upper Body Framing for crops that keep the legs out of frame.

Anchor: https://berrys.io/research/ai-model-pose-library#finding-framing-mix
Denominator: 100 poses by framing series

### 50 of 100 poses (50%) are seated, kneeling, reclining or crouching, so a bag, shoe or torso-up SKU has 50 ground-level posing options before a single standing shot is considered.

Anchor: https://berrys.io/research/ai-model-pose-library#finding-below-waist-share
Denominator: 50 non-standing poses / 100 poses

## Posture families

| Posture family | Poses | Angles | Share of poses | Best for |
| --- | --- | --- | --- | --- |
| Hand Gestures | 27 | 135 | 27% | Torso-up crops, hold-the-product frames, accessories and small electronics |
| Seated | 23 | 115 | 23% | Bottoms, footwear on a low stool, home and lifestyle SKUs |
| Standing | 17 | 85 | 17% | Full-length apparel, outerwear, bags, full-body catalogue covers |
| Kneeling | 12 | 60 | 12% | Denim, activewear, mid-length dresses, ground-plane product placement |
| Floor Recline | 9 | 45 | 9% | Swimwear, loungewear, bedding and home textiles |
| Crouch | 6 | 30 | 6% | Sneakers, sports accessories, streetwear lookbooks |
| Forward Lean | 6 | 30 | 6% | Torso-up beauty and skincare, shelf and counter contexts |

## Asset coverage per angle

| Asset per camera angle | Count | Coverage |
| --- | --- | --- |
| ControlNet depth map (PNG) | 500 | 100% |
| Line-art mannequin render (PNG) | 500 | 100% |
| Thumbnail (JPEG) | 500 | 100% |
| Written pose brief | 426 | 85.2% |

## Camera angle set

- Front: 100 angle records
- Top-Down: 100 angle records
- Low-Angle: 100 angle records
- Side: 100 angle records
- 3/4 View: 100 angle records

Every pose carries all 5 views. That is the property that matters for catalogue work: a garment can be rendered as a 5-view set from one pose record, and the depth map for each view pins limb position so drape and hem line stay comparable between frames.

## Method

- Source: the published pose library JSON at https://berrys.io/poses-full.json (mirrored from lib/poses_full_sfw.json), 100 pose records.
- Unit of counting: the pose record (100) and the camera-angle record (500). One pose record always contains five angle records.
- "Written pose brief" means the `brief` field on an angle record is non-empty after trimming. Empty strings are counted as missing, not inferred from the pose name.
- "Distinct briefs" deduplicates identical brief text across angle records, which is why 426 briefed records reduce to 98 unique descriptions.
- Shares are rounded to one decimal place. Comparisons within one pose of each other are treated as level, not as a difference.
- Excluded: no generated images, no model output, no platform compliance data. This study counts the reference dataset only.
- Refresh: recomputed whenever the dataset is regenerated. Last updated 2026-11-06.

## How to use the data

1. Pick a family from the posture table by product type, then note the pose code (for example P001).
2. Fetch the depth map for the view you need (`/for/poses/img/depth/full_single_<code>_<angle>.png`) and load it as the pose reference on the Berrys canvas next to your File Node packshot.
3. Read the written brief when one exists; where the brief is empty (74 of 500 records), describe the geometry from the depth map instead of guessing from the thumbnail.
4. Keep the Brand System Node upstream so lighting, palette and lens character stay identical across the 5 views.

## Suggested citation

Berrys AI (2026-11-06). AI Model Pose Library Statistics 2026: 100 Poses, 500 Angles. https://berrys.io/research/ai-model-pose-library

## FAQ

### How many poses are in the Berrys pose library?

100 poses, published as 500 camera-angle records (5 angles per pose), with 500 ControlNet depth maps.

### Do all poses include a ControlNet depth map?

Yes. 500 of 500 angle records (100%) ship a depth map, together with a line-art render and a thumbnail.

### How many poses have a written prompt brief?

426 of 500 angle records (85.2%) carry a written brief averaging 17.5 words. The remaining 74 are visual-only.

### Which pose family has the most options?

Hand Gestures, with 27 poses and 135 of 500 angle records.

## Related

- [Pose library gallery](https://berrys.io/for/poses)
- [Pose dataset (JSON)](https://berrys.io/poses-full.json)
- [Marketplace image compliance checker](https://berrys.io/for/tools/marketplace-image-checker)
- [Research index](https://berrys.io/research)
- [AI usage policy](https://berrys.io/ai.txt)