State of AI Fashion Imagery 2026
How brands actually use AI to make fashion imagery - from 8,300+ on-model images, 2,900+ products and 250+ AI models on Flashlook - and how the AI image models behind the category compare in LMArena's blind human votes.
Six numbers worth knowing
85%
of on-model images made on Flashlook use a 4:5 portrait frame; square (1:1) is under 2%.
72%
of images are shot on ready-made library models rather than a brand's own custom model.
5.6
images per product on average - brands shoot several angles and scenes per SKU, not one hero.
40%
of categorized products are tops - the most-shot category, ahead of trousers and knitwear.
#1
GPT-Image-2.5 Sunburst (OpenAI) leads LMArena's photorealistic text-to-image board and also single-image editing; the best openly licensed model on the photorealistic board (Cosmos3-Super-Text2Image (Agentic)) ranks #33.
8,300+
on-model images generated on Flashlook between June 2026 and November 2026.
On-model images generated per month
Stills made on Flashlook. *Month to date.
- Jul: 239
- Aug: 2,638
- Sep: 1,458
- Oct: 2,752
- Nov: 1,202 (month to date)
Image format
Share of on-model stills by aspect ratio.
What gets shot
Share of categorized products by category.
Library models vs custom models
Share of on-model stills by the kind of AI model used.
Library: 95 ready-made models anyone can shoot with. Custom: models a brand builds for itself.
Who leads AI image generation for commercial work
From LMArena's public leaderboards (published Nov 17, 2026), ranked by blind human votes.
Text to Image · Photorealistic & Cinematic Imagery
Full board →| Rank | Model | Score | Votes |
|---|---|---|---|
| 1 | GPT-Image-2.5 SunburstOpenAI · Proprietary | 1434 ±8 | 23,046 |
| 2 | GPT-Image-2.5 FlareOpenAI · Proprietary | 1417 ±7 | 22,431 |
| 3 | GPT-Image-2.6 Sunburst (Medium)OpenAI · Proprietary | 1411 ±14 | 2,331 |
| 4 | MAI-Image-2.7Microsoft AI · Proprietary | 1389 ±20 | 1,032 |
| 5 | GPT-Image-2.6 Flare (Medium)OpenAI · Proprietary | 1388 ±14 | 2,234 |
Best openly licensed model: Cosmos3-Super-Text2Image (Agentic) (OpenMDW-1.1) at #33.
Text to Image · Product, Branding & Commercial Design
Full board →| Rank | Model | Score | Votes |
|---|---|---|---|
| 1 | GPT-Image-2.5 SunburstOpenAI · Proprietary | 1421 ±7 | 21,834 |
| 2 | GPT-Image-2.6 Sunburst (Medium)OpenAI · Proprietary | 1396 ±15 | 2,079 |
| 3 | GPT-Image-2 (Medium)OpenAI · Proprietary | 1386 ±5 | 40,502 |
| 4 | GPT-Image-2.5 FlareOpenAI · Proprietary | 1383 ±7 | 21,532 |
| 5 | MAI-Image-2.7Microsoft AI · Proprietary | 1376 ±19 | 1,098 |
Best openly licensed model: Cosmos3-Super-Text2Image (Agentic) (OpenMDW-1.1) at #35.
Image Edit · Single Image · Overall
Full board →| Rank | Model | Score | Votes |
|---|---|---|---|
| 1 | GPT-Image-2.5 SunburstOpenAI · Proprietary | 1519 ±4 | 140,743 |
| 2 | GPT-Image-2.5 FlareOpenAI · Proprietary | 1484 ±4 | 137,820 |
| 3 | GPT-Image-2.6 Sunburst (Medium)OpenAI · Proprietary | 1481 ±7 | 16,881 |
| 4 | MAI-Image-2.7Microsoft AI · Proprietary | 1476 ±11 | 3,007 |
| 5 | GPT-Image-2 (Medium)OpenAI · Proprietary | 1462 ±3 | 371,930 |
Best openly licensed model: Qwen-Image-Edit (Apache 2.0) at #42.
Image Edit · Multi Image
Full board →| Rank | Model | Score | Votes |
|---|---|---|---|
| 1 | GPT-Image-2.5 SunburstOpenAI · Proprietary | 1527 ±5 | 70,812 |
| 2 | GPT-Image-2.5 FlareOpenAI · Proprietary | 1475 ±5 | 69,942 |
| 3 | MAI-Image-2.7Microsoft AI · Proprietary | 1458 ±10 | 3,647 |
| 4 | GPT-Image-2 (Medium)OpenAI · Proprietary | 1455 ±3 | 167,372 |
| 5 | GPT-Image-2.6 Flare (Medium)OpenAI · Proprietary | 1442 ±7 | 10,635 |
Best openly licensed model: Qwen-Image-Edit (2511) (Apache 2.0) at #37.
What the wider market says
Selected third-party figures - see all of them, with sources, on the statistics page.
61%
61% of shoppers consider product images and videos the most important product page element for completing a purchase, compared with 60% for pricing and availability and 57% for ratings, reviews and UGC.
24.52%
The US retail return rate on online sales was 24.52% in 2024, versus 8.72% for in-store purchases and 13.21% overall ($685B).
$50–$350
Per-image product photography pricing can land anywhere between $50 and $350, and photographers may charge $500 to $3,000 per day.
75%
75% of Shopify store owners reported using AI tools for tasks like content generation, data analysis and store operations; content generation was the most common use case (69%).
60%
60% of US online shoppers say knowing a product image is AI-generated reduces their purchase confidence, and 17% say they would not feel confident enough to buy at all; 18% say it doesn't affect them.
48.6%
48.6% of marketers rank short-form video in their top three formats for ROI, the highest of any media format.
Methodology
- Flashlook data - anonymized aggregates of on-model stills generated on Flashlook from Jun 29, 2026 to Nov 16, 2026, including images made by Flashlook's own team. Totals are rounded down; no customer, brand or user is identified.
- Categories - share of products that have a category set; products filed as "other" or uncategorized are excluded.
- AI model rankings - LMArena's public image leaderboards (CC BY 4.0), published Nov 17, 2026. Flashlook doesn't run those votes or change the scores.
- Industry figures - third-party statistics, each checked against its original source, linked inline.
Flashlook (2026). State of AI Fashion Imagery 2026. https://flashlook.ai/reports/state-of-ai-fashion-imagery-2026
Text and data licensed CC BY 4.0 - reuse freely with a link back. Arena rankings © LMArena, CC BY 4.0.
Report data (JSON)Make the images in this report
Turn product photos into on-model images and video on library or custom AI models. Free credits on signup.


