sequenced.ai
Articles/Content & marketing/Blueprint//7 min read

How Cloudinary extends product photography without losing the original

Cloudinary can turn one photograph into a family of delivery assets. Generative fill adds new scenery, so the publishing decision includes what an editor approves and what each variant costs.

By Sequenced deskAI-assisted, source-led · how we work
Visit Cloudinary website ↗
FillAI operationExtends the image into a padded region.
SeedVariationSelect another background for the same prompt.
EagerPreparationPrepare derived assets before delivery traffic arrives.
CreditsBilling poolProcessing shares an allowance with storage and bandwidth.
Cloudinary mark
Cloudinarycloudinary.com · independent research

Represent this company? Verify your work email to access its workspace, or send the desk a factual correction.

The hero slot on a shop's home page is wide, the photograph of its new coffee grinder is tall, and a conventional crop cuts off the handle. Cloudinary offers another option. Its generative fill can extend the surrounding scene while keeping the existing photograph in the frame. For a retailer, the interesting part comes after that first plausible image. Someone has to approve what was invented, reproduce the approved variant and deliver it reliably to the next visitor.

In brief
  1. 01The platform Cloudinary manages image and video assets and delivers transformed versions. Generative edits can run as part of the same media pipeline.
  2. 02The constraint Fill uses padding on an eligible opaque image. The placement of the source photograph determines where the model adds pixels.
  3. 03The accounting A generated fill carries a special transformation count. Storage and delivery draw on the same monthly credit pool as processing.

01 / ProductA media pipeline that can create as well as resize

In an ordinary resize, the destination contains pixels derived from the original scene. Extending an image with generated content gives the editor a new responsibility. A longer countertop is plausible, but an invented power cord or an extra accessory might tell the customer something false about the item. Cloudinary's generative transformation overview distinguishes filling, recoloring, object removal, replacement and other operations. They solve different visual problems and deserve different approval criteria.

The fill guide describes the mechanics. Specify a padded size or aspect ratio and use the generative fill background option. Positioning the source with gravity controls which side needs extra material. A prompt can guide that new area. A seed can select a different outcome, and reproducing the chosen result also depends on preserving the preceding transformation parameters. Saving a prompt by itself leaves part of the recipe out.

A delivery platform makes that distinction operational. A retailer might need a wide banner, a narrow card and a small thumbnail from the same photo. Some fit the original perfectly and need only a normal resize. Others introduce a new patch of scene. A catalogue that records which variants contain generated material lets the editor direct scrutiny toward those new pixels and keep routine delivery work simple.

02 / AudienceWho gains from treating an edit as a repeatable instruction

A merchandising team with a regular stream of new products often has excellent originals and too few shots for every layout. A developer can define a limited set of useful output shapes, while the art director decides which subjects can tolerate generated surroundings. The arrangement is particularly attractive when a change to the storefront would otherwise send hundreds of photos back through a manual export process.

The case is weaker for product details where the surrounding region forms part of the evidence. A photograph showing how a mounting bracket fits against a wall conveys dimensions and clearances. Extending that wall may be harmless in a decorative banner yet misleading in an installation guide. Choose a policy by image role. “Campaign background” and “technical documentation” can justifiably have different rules even when both display the same product.

03 / WorkflowA proposed workflow for the grinder campaign

Start with an uploaded, opaque studio photograph that has a clean outline and enough space around the product. Preserve that file as the reference. Define the banner's crop geometry, put the grinder toward the side away from the intended headline, and ask for a background consistent with the existing surface. Produce a plain padded version as a comparison. The exercise tests whether generation earns its place beside a simpler layout that the art director can already trust.

Review the result at the actual banner size and inspect details at full resolution. Look where newly invented counter space meets a shadow, where a cable touches the frame and where fine lettering approaches the edge. Check the product against the original. If the surface suggests that an accessory is bundled, reject the variant even if it looks photographic. Record the source asset, transformation sequence and chosen seed with the approval.

Next prepare the chosen derivative before a launch preview. The eager transformation guide explains how to generate defined variants ahead of the first delivery request and how asynchronous completion can be reported. This matters because the fill guide documents a possible HTTP 423 response while a derivative is being generated. A draft storefront can keep the approved conventional crop available until the generated asset is ready.

Suppose the art team wants the same composition at a smaller size. Try reducing the approved output after the fill step. Changing the original padding geometry before generation changes the input seen by the model. Those are different approval paths. Name the variants in a way that lets a reviewer distinguish scaling an accepted composition from commissioning a new one. The first operation often needs a legibility check, while the second calls for another review of the depicted scene.

04 / PricingCounting fills, deliveries and storage separately

The Image and Video API pricing page describes monthly credits that can be spent across transformations, managed storage and delivery bandwidth. Its US-dollar display lists Free with 25 monthly credits, Plus with 225 and Advanced with 600. The monthly billing view lists Plus at $99 per month and Advanced at $249. The site also offers lower monthly equivalents on yearly billing. These are platform allowances and should be read alongside the individual effect's processing count.

PlanMonthly priceMonthly credits
Free$025
Plus$99225
Advanced$249600

Cloudinary API plans in US dollars on monthly billing, consulted 19 November 2026. Published plans and transformation counts.

In the transformation count reference, generative fill counts as 50 transformations. The plan page equates one credit with 1,000 transformations. As a worked processing-only example, 20 distinct fills at that stated count use 1,000 transformation units, equivalent to one credit. This arithmetic excludes other operations, stored originals and derivatives, and delivery traffic. Twenty people opening an already generated identical rendition is a different cost pattern from creating 20 different scenes.

For a real estimate, count versions per product as well as products per release. Record the number of artistic variations tried, the approved versions retained, the delivery formats requested and how much data the campaign serves. A widely visited campaign might use a modest number of generated backgrounds and substantial image bandwidth. A small creative team exploring many alternatives might see the reverse. A single “cost per image” borrowed from either example can mislead the other.

05 / DistinctionsChoosing between an editing workspace and a delivery service

The Adobe Blueprint provides a useful adjacent case when a designer needs to compose and retouch a few high-value images individually. Compare the effort required to perfect one difficult photograph with the effort required to deliver new variants of hundreds of approved ones. A team can use an editor to establish a master and still choose a delivery service to adapt it across placements.

The Photoroom Blueprint is relevant when the buyer starts from product cutouts and ready-to-publish catalogue imagery. Cloudinary's particular attraction here is the relationship between the existing asset, a reproducible transform and the URL that serves its result. Evaluate the complete journey from original to approved delivery. An impressive sample does little for a retailer if staff cannot tell which version is on a product page.

06 / QuestionsWhich images need a different route

Cloudinary's fill guide excludes transparent, animated and fetched images. A pipeline that accepts a supplier's remote image URL as a fetched resource should account for that distinction before promising generative banners. A transparent pack shot also needs a deliberate composition decision. Test eligible uploaded originals first and keep the approved flat-background treatment for a product that has no suitable source photograph.

Background removal is a separate operation with a commercial transition of its own. Cloudinary says accounts created after 1 February 2026 cannot subscribe to the older AI Background Removal add-on and directs customers to the transformation version. Existing subscribers have a migration path with billing-cycle considerations. A new team reading older examples should design around the current transformation and its count. Background removal availability notice

07 / DecisionChoose the image policy before scaling the catalogue

01

A retailer with a strong photo library

Try one campaign with a source that needs extra space but already shows the entire product. Compare ordinary padding with several generated results at the true delivery size.

Approve the composition and preserve its exact instructions.
02

A team publishing technical product imagery

Identify which images communicate dimensions, fit or included parts. Keep their documentary source intact and limit generated scenes to placements where they cannot change those claims.

Give editors a clear distinction between evidence and illustration.
03

An operator estimating a catalogue rollout

Count new derivatives, art-direction alternatives, managed storage and expected delivery. Generate the approved assets early and exercise the fallback before the launch.

Base the budget on assets created and traffic actually served.
What should we explore next?

A business worth understanding.

Suggest your business or one you find interesting. Tell us what you want to understand about its product, positioning, design or workflows.

Suggestions are free. Selection and publication stay with the desk.

Sources
Filed under Content & marketingCompany CloudinaryNot affiliated with CloudinaryRequest a correctionRequest a refresh by email

Continue reading

All in this category