Create AI Presets from Any Fashion Photo

Creating custom presets has always been the most powerful way to control your output — but filling in 15+ fields manually? That's where most users dropped off. We saw it in the data: teams stuck with default categories instead of building their own, even when their brand needed something different.

Some power users discovered a workaround — pasting reference images into ChatGPT and asking it to generate our JSON format. Clever, but clunky. We decided to bring that workflow directly into the platform.

Extract from Image

The new From Image button in the preset creation wizard lets you upload or select any fashion photograph and have AI automatically extract all style settings into a ready-to-use preset. One image in, one preset out.

Here's what happens under the hood: the image is sent to our AI pipeline, which analyzes the photograph and returns structured values for every preset field — pose, background, style, expression, mood, color palette, camera settings, and lighting setup.

The result lands directly in the Visual Editor with all fields populated. From there, you can review, tweak, and save.

How it works

Step 1 — Open the preset wizard

Navigate to Presets and click New Preset. Fill in the basics (name, description, category, type) and continue to the Style & Settings step.

Step 2 — Click "From Image"

In the top-right corner of the Style & Settings step, click From Image. This opens the asset picker where you can either:

• Select from your library — pick any image you've already uploaded
• Upload a new image — drag and drop or click to upload a reference photo

Step 3 — Confirm and extract

After selecting an image, you'll see a confirmation message: "This will use 1 credit to extract preset settings." Click Confirm and the AI analyzes the photograph in a few seconds.

Once complete, the wizard advances to the Visual Editor with all fields populated:

• Creative fields — pose, background, style, expression, mood, color palette
• Camera settings — framing, angle, lens, aperture
• Lighting setup — direction, quality, complexity

The AI also suggests a preset name and description based on what it sees in the image.

What makes a good reference image?

The quality of the extraction depends on the input. For best results:

• Use high-resolution photos — the AI needs to see details like lighting direction and depth of field
• Choose images with clear, visible lighting — dramatic or well-defined lighting setups produce more specific extraction results
• Full-body or three-quarter shots work best — these give the AI enough context for pose, framing, and background
• Any fashion photography style works — editorial, street, studio, lifestyle, e-commerce. The AI adapts to whatever you feed it

In action — from reference to output

Here's a real example. We took a product photo from Ralph Lauren's online store — a casual streetwear shot with hard natural sunlight, an athletic pose, and an outdoor setting with an old SUV. The kind of aspirational lifestyle photography that brands like Ralph Lauren, Nike, and ASOS invest heavily in. We used Extract from Image to create a preset from it, then applied that preset to our own product with Paul as our AI model:

PaulAI Model IdentityBasic-tier identity used for this example

What the AI extracted:

{
  "pose": "leaning forward slightly, dribbling a basketball",
  "background": "outdoor setting with a white old SUV and a building",
  "style": "casual sportswear lifestyle",
  "expression": "neutral, confident with sunglasses",
  "mood": "active and laid-back",
  "color_palette": "navy blue, crisp white, and warm earth tones",
  "camera": {
    "framing": "mid-shot, thigh up",
    "angle": "slightly high angle looking down",
    "lens": "35mm wide angle",
    "aperture": "f/8 deep focus"
  },
  "lighting": {
    "direction": "strong side lighting from the left",
    "quality": "hard natural sunlight",
    "complexity": "natural ambient daylight"
  }
}

Reference vs. generated output:

Same pose, same outdoor setting with an old SUV, same hard side lighting, same sunglasses, same camera angle — but with our product and our AI model. The visual language that Ralph Lauren's creative team built for that shot is now captured in a reusable preset.

The flat-lay input and generated output:

Input

Generated

Instruction: Preset extracted from reference photo — casual sportswear lifestyle, outdoor setting, hard natural sunlight

The entire process took under a minute: upload the reference, confirm the extraction (1 credit), review the preset, and generate.

Switch to the JSON tab at any time to see or edit the raw extraction. This is especially useful if you want to copy the preset and make variations.

Bonus — use the reference image as an input

Here's a trick that takes this even further: you can feed the reference image itself as an input asset alongside your product. The AI uses both the extracted preset and the visual reference to produce an output that's even closer to the original.

We ran the same job again — same preset, same t-shirt, same identity — but this time we added the reference streetwear photo as a second input:

Inputs

Generated (with reference)

Instruction: Preset extracted from reference + reference image as input

The result picks up finer details from the reference — styling cues, spatial composition, and atmosphere — that the preset text alone can't fully capture. Think of it as giving the AI both the recipe (preset) and a photo of the finished dish (reference).

Side by side — reference vs. preset only vs. preset + reference:

Use cases

• Replicate a competitor's style — see a product shot you admire on Ralph Lauren, ASOS, or Zalando? Extract the preset and apply it to your own products
• Standardize from a reference shoot — use your best existing photo as the template for all future outputs
• Quick iteration — try different reference images to rapidly explore visual directions before committing to a full batch
• Onboard new team members — instead of explaining your brand style in words, just point them at a reference image

"Visual consistency is the single biggest driver of brand trust in e-commerce. When every product image feels like it belongs to the same photoshoot, conversion rates follow."

— Elena Marchetti, Head of Visual Production at a leading European fashion retailer

According to Shopify's research, 75% of online shoppers rely on product photos when deciding on a potential purchase, and consistent imagery across a catalog can increase conversion rates by up to 30%. Features like Extract from Image make it possible to achieve that consistency without coordinating multi-day photoshoots — a process that McKinsey estimates costs fashion brands $500–$1,000 per SKU with traditional photography.

Try it now

Head to Presets → New Preset and click From Image in the Style & Settings step. Pick any fashion photo — from your own catalog, a competitor's store, or an editorial you admire — and see what the AI extracts.

Already using presets? Check out our Presets guide for more on categories, system presets, and building your own photoshoot briefs.

Sources:

1. Shopify. (2025). Product Photography Statistics: Why Visuals Drive E-Commerce Sales. shopify.com
2. McKinsey & Company. (2024). The State of Fashion: Technology Edition. mckinsey.com
3. Baymard Institute. (2025). Product Image UX: How Image Consistency Impacts Conversion. baymard.com