Glossary of fashion image production

Definitions of the metrics, production terms and failure modes of generated fashion imagery: garment fidelity, character consistency, acceptance rate, Delta E and more.

Metrics

Garment fidelity: Garment fidelity is how closely a garment in a generated or edited image matches the real product in silhouette, length, construction, colour, print, logos, hardware, texture and fine detail. The garment fidelity rate is the share of reviewed images that pass on every one of these attributes.
Character consistency: Character consistency is the degree to which a model looks like the same person across a set of generated fashion images: same face, skin tone, body proportions and hair. It is tracked as the character drift rate, the share of reviewed images rejected for an identity change.
Acceptance rate: Acceptance rate is the share of generated fashion images that pass review and can be delivered without changes. Its inverse, generations per accepted image, shows how much production and review effort one usable image costs.
Cost per accepted image: Cost per accepted image is the total cost of producing images, including generation, rejected attempts, review time, corrections and set-up, divided by the number that pass review. It makes generated imagery comparable with a photoshoot's cost per finished image.
Delta E (colour difference): Delta E (ΔE) is the standard measure of colour difference: one number for how different two colours look. In product imagery it checks a garment's colour against its packshot and a background against its target; as a rule of thumb, a difference under 2 is hard to see without close inspection.

Production terms

Managed fashion image production: Managed fashion image production is a service in which a partner generates on-model images for a brand's product pages and campaigns from its product references, using models and shot rules the brand has approved, and delivers only images that passed the partner's review.
PDP-to-campaign production: PDP-to-campaign production means making lookbook, lifestyle or campaign images from the product references and approved models already used for product pages, instead of organising a separate shoot for each channel.
Product references: Product references are the images that define how a garment looks, such as packshots, detail shots and back views. In generated fashion imagery they are the source of truth for colour, shape and detail.
Approved model identity: An approved model identity is a reusable model, generated or based on a consenting real person, that a brand has reviewed and signed off as a set of reference images covering every angle it shoots, so the same person appears across products and seasons.
Deep fake (EU AI Act): Under the EU AI Act, a deep fake is AI-generated or manipulated content that resembles existing people, objects, places or events and would falsely appear authentic. A photorealistic generated fashion model likely qualifies, which triggers a visible disclosure duty since 2 August 2026.

Failure modes

Logo distortion: Logo distortion is a failure in generated fashion imagery where a logo, badge or lettering on a garment is redrawn: letters turn into near-letters, shapes change, or a mark moves or disappears.
Print drift: Print drift is a failure where a garment's print or pattern changes in a generated image: a different scale, a shifted placement, a simplified repeat, or a pattern replaced by a similar one.
Colour shift: Colour shift is a failure where a garment in a generated or retouched image differs in hue, saturation or lightness from the real product. In a 2023 Coresight Research survey, US apparel retailers named colour as the second most common reason for returns, after size and fit.
Construction drift: Construction drift is a failure where a garment's build changes in a generated image: pockets, buttons, seams, panels, collars or closures appear, disappear or change shape compared with the real product.
Texture hallucination: Texture hallucination is a failure where a generated image shows a fabric surface the product does not have: a cable knit rendered smooth, a flat weave given a sheen, or a denim wash invented instead of copied.
Reference confusion: Reference confusion is a failure where attributes from one reference image bleed into another element of a generated image, such as the colour of a second garment appearing on the first.
Anatomy errors: Anatomy errors are failures in generated fashion imagery where a body is rendered wrongly: extra or missing fingers, fused hands, a foot or ankle turned the wrong way, or limbs out of proportion.
Background contamination: Background contamination is a failure where an e-commerce background that should be clean white or a set grey shows shadows, gradients, stray objects or a colour cast, measured against the target colour with Delta E.