Photoroom

02 / IDENTITY

About the company

Key company facts

Founders and current leadership

Founding story and history

PhotoRoom’s initial commercial attraction was concrete rather than speculative: merchants photographing stock in kitchens, closets and makeshift studios needed clean, sale-ready photos. Background removal was an easy entry point; generated scenes and repeatable editing made the software more useful to brands and platforms. [8] [29]

THE BUSINESS SHIFT

Why e-commerce became central

Online sellers need many consistent pictures—not just one attractive edit. PhotoRoom’s proposition evolved from removing backgrounds to assembling complete, repeatable image workflows, including product-preservation checks that matter when buyers expect the physical item to match the image.

Editorial illustration: Photoroom official website. [1] [21]

Ownership and corporate structure

The corporate legal notice identifies Photoroom SAS as the French website operator. The company remains a venture-backed private business; public disclosure does not identify a controlling parent, shareholder percentages or complete subsidiary structure. Its 2025 GenerateBanners acquisition is confirmed. It would be incorrect to treat Balderton, Aglaé or Y Combinator as the operating parent merely because they invested. [2] [5] [9]

03 / PRODUCT ECOSYSTEM

What Photoroom actually sells

The core product is no longer one background-removal button. Photoroom combines AI editing, photo staging, catalog tooling, creative generation and production-grade APIs, packaged differently for individual sellers, teams and enterprise buyers. [1] [10] [11] [20]

Current products and capabilities

Product names and classifications follow 2025–2026 official pages. Access varies by device, plan, AI credits and beta rollout; this overview is not an entitlement guarantee. Product Fixer should not be interpreted as perfect automatic fidelity. [9] [10] [11] [20] [21] [40]

How the main workflows work

These are representative editorial workflows based on official descriptions—not screenshots of the user interface. Background removal edits are distinct from wholly synthetic text-to-image generation. [12] [20] [44] [45]

Platform availability and users

Web + iOS + Android: broad creator/seller product; a login and a Space are used for work. Shopify: a directly connected listing tool, not just an export format; publishing depends on plan. API: keys and separate usage billing. Capture One: embedded integration in supported Studio plans. A standalone general-purpose Windows desktop app was not confirmed from official sources. [1] [20] [27]

Industries with documented relevance

Ecommerce and marketplaces remain core, with explicit use cases in fashion and apparel, beauty, home goods, food and delivery, furniture resale, automotive, advertising, social commerce and SaaS. The presence of a use-case page does not automatically establish a paying customer or formal partnership in that vertical. [1] [23] [25] [26]

04 / MODELS & ENGINEERING

How its AI works—and what is genuinely confirmed

Photoroom is not merely a graphical layer over one external foundation model. Its engineering disclosures document proprietary segmentation and diffusion work, open research artifacts, and use of third-party image models for specific features. The exact production model behind every button is not publicly mapped. [16] [17] [18] [19]

Confirmed technology and named models

General technical explanation (not a claim about its exact architecture)

Segmentation locates product pixels and produces a mask, often refined through matting and edge preservation. Inpainting rebuilds selected missing regions. Diffusion image generation transforms noise into content guided by a prompt or reference. Outpainting extends the frame. Relighting and shadow synthesis attempt to match illumination and contact shadows. Modern products may chain these operations and add quality checks. These are general categories, not confirmation of PhotoRoom’s internal implementation for each feature.

Research and engineering evidence

Public resources: PRX GitHub repository · Dataroom code · PRX Pixel weights. The precise on-device versus cloud routing for each present-day tool, and a complete production model registry, are not publicly disclosed. Enterprise compute location and hosting controls are described separately in security materials. [17] [28] [15]

Product fidelity: a hard limit of generative images

Photoroom’s August 2026 internal evaluation tested thousands of generated product images and reported that even strong editing models could modify item details, including logos and text. The company markets Product Fixer, automated QA and an Enterprise Guarantee to address these failures. These internal benchmark results were company-reported, not independently replicated; they are evidence of a problem and an attempted mitigation, not proof of perfect output accuracy. [21]

For marketplace products, human review remains prudent for ingredient lists, labels, SKU colors, jewelry details, garment stitching and condition marks. Even when the photo background looks better, the product must remain a truthful representation.

05 / COMMERCIAL MODEL

Plans, pricing and monetization

The business monetizes editor subscriptions, credits/exports, metered APIs, and negotiated enterprise workflows. The published consumer editor prices below are USD monthly equivalents billed annually from the website; app-store, geography and month-to-month prices can differ. [10] [11]

PRICING CHECKED · 16 OCTOBER 2026 · WEB PRICING DISPLAY · USD

The official help center describes a Space-based billing model, where a subscription covers a Space rather than charging per invited seat, and says up to 49 additional people can join a Space. Tool-specific credits and fair-use limits still apply. [10] Billing for Spaces

Developer API: capabilities, price and limits

API access is billed separately from normal app subscriptions unless under an eligible negotiated enterprise arrangement. The API dashboard provisions credentials; requests use an x-api-key header. Basic is for removal and straightforward image cleanup; Plus enables more extensive editing and generated scenes. [11] [12]

Pricing caveat: Photoroom’s public API FAQs state $0.02 Basic and $0.10 Plus cross-plan call pricing; contracted rates, other creative operations and volume discounts may differ. Plus input commonly supports PNG/JPEG/WebP/AVIF, up to 30MB and a 5,000px longest-side processing limit; Basic input permits HEIC and up to 50MB/6,000px longest side according to official docs. See current API terms before deployment. [11] [12]

How the API business fits the strategy

Embeddable image processing can reduce seller friction for a marketplace (see Depop), standardize merchant catalogs (Wolt), and automate large internal DAM/PIM pipelines (Decathlon). This creates recurring usage at the business-process layer, distinct from subscriptions sold to individual editors. The percentage of total company revenue attributed to APIs has not been publicly disclosed. [23] [24] [25]

Confirmed revenue channels

Confirmed: subscription plans, usage-based API plans, annual enterprise commitments and contracted automation. Likely but unverified: relative revenue split among consumer, enterprise and API; amounts from services or specialized onboarding. Avoid assuming every integration generates direct commission revenue. [10] [11]

06 / CAPITAL & SCALE

Funding, investors and financials

The last verified public financing is the February 2024 Series B. A 2024 figure must not be repackaged as a fresh 2026 capital raise or current valuation. [5] [6]

Company and investor statements place cumulative funds at $64 million as of the 2024 Series B. Simple arithmetic on independently reported historical rounds can differ slightly because announcements round figures and may omit minor financings; the company’s own total is preferred. [4] [5] [30]

Major investors and ownership boundaries

Verified named investors: Balderton Capital, Aglaé, Y Combinator and Adjacent (in the public financing sources). Historical business reporting also names Kima Ventures, FJ Labs, Meta and individual angels including Yann LeCun, but does not establish check size, board seat or participation in every round. Shareholdings and ultimate control are not publicly disclosed. [4] [5] [6]

Valuation and revenue

Downloads, usage and company size

07 / MARKET PROOF

Named customers and actual outcomes

Photoroom’s customer library now includes documented marketplace and enterprise deployments. The results below are customer stories published by Photoroom, with client quotes or described tests where available; no independent audit of operational metrics was found. [22] [23] [24] [25] [26]

What these studies do—and do not—prove

These examples establish direct working relationships, implementations and attributed results. They do not demonstrate identical savings or conversion improvements for an unrelated retailer. The Vestiaire experiment is a more useful comparative design than a simple before/after testimonial, but the data were still released through a vendor announcement. [22]

Other companies named in PhotoRoom customer materials include Mercari, GoodBuy Gear, Valuence Japan, Goodwill and Warner Bros. They are documented case-story subjects; a customer logo alone would not have been sufficient. See official customer story library.

Partnerships, integrations and acquisitions

Reference integration directory: Photoroom Integrations. No unverified reseller, app, cloud or marketplace arrangement is implied by platform compatibility.

08 / TRUST, RIGHTS & RISK

Privacy, training, ownership and security

This is the part of the company profile where technical differences between app uploads, API processing, retained projects and model-training permissions matter most. Its public documents are not fully aligned on API training defaults. [13] [14] [15] [37]

Privacy and training: the primary-source conflict

Storage, deletion, location and controls

Ownership and commercial use

PhotoRoom’s July 2026 Terms say customers retain ownership of their Customer Content and that Photoroom does not claim ownership of it. The Terms also grant the company broad licenses for processing, provision of services and stated service improvement. Possessing an edited image does not automatically clear third-party trademark, copyright, likeness or advertising rights. AI-generation output rights may depend on the specific tool, terms, input material and governing law. [14]

Security and AI governance

Photoroom says its SOC 2 Type 2 attestation is scoped to the API, with an April 2026 report accessible through its Trust Center. It describes TLS 1.2+ in transit, encryption at rest through infrastructure providers, access controls, subprocessor information and an incident-response process. Those are company-published representations; this research did not independently review the underlying audit report. [15]

Its responsible-AI statement says it trains internal models on selected public datasets, licensed partner datasets, synthetic data and consenting user content. It acknowledges third-party models do not necessarily follow the exact governance of wholly in-house systems. [16]

For product imagery, users should avoid generating misleading packaging, changing product defects, misrepresenting ingredients or producing artificial customer endorsements. The absence of a visual defect is not proof a generated picture is accurate.

09 / ANALYSIS

Reviews, strengths and limitations

Photoroom occupies a useful middle ground: quicker than deep manual compositing for routine selling images, but not a guarantee that generated scenes will preserve every physical detail.

Documented strengths vs genuine constraints

What users say—treated as sentiment

Independent seller comments praise quick cutouts and useful speed savings, while others report credit confusion, regressions, occasional background errors and cancellation/support frustration. Reddit is self-selecting; these anecdotes are not representative quality scores, a confirmed service-wide defect rate or an independently measured satisfaction benchmark. [38] [39]

For purchasing decisions, test your own fragile product samples—glass, hair, jewelry, transparent packaging, printed labels—before selecting a paid production tier.

Who benefits most?

10 / COMPETITIVE LANDSCAPE

Where Photoroom sits in the market

Photoroom’s closest competition is not a single category: Pixelcut overlaps heavily on AI product visuals; remove.bg specializes in cutouts; Canva and Adobe provide broader design ecosystems. [35] [36] [50] [52]

Capabilities verified through the providers’ live feature descriptions, not uniform hands-on benchmarks or price-matched trials. Pixelcut’s published API includes removal, background generation and try-on. Canva lists bulk background removal; Adobe lists generative fill. [35] [36] [50] [51] [52]

11 / KNOW BEFORE YOU BUY

Frequently asked questions

Concise answers based on the cited sections above. Open any question to read the answer.

12 / EDITORIAL METADATA

SEO publishing brief

Recommended search title

Photoroom Company Profile 2026: Founders, AI, Funding & API

Alternative title directions

• Photoroom: AI Product Photography Company, Technology & Pricing
• Photoroom Founders, Funding, Models, Customers and Competitors
• Photoroom Company History: From Background Removal to Enterprise AI
• Photoroom API & AI Photo Editor: Company Profile and Valuation

Meta description

Explore Photoroom’s founders, funding, AI photo tools, pricing, API, customers, privacy, research models and competitors in this 2026 company profile.

Suggested slug

/companies/photoroom/

Short company description

Photoroom is a French AI image-editing company best known for removing photo backgrounds and helping sellers create product photography. Founded in 2019, it now offers generative product scenes, fashion imaging, batch editing, catalog tools, developer APIs and enterprise visual automation. Its products serve ecommerce brands, online sellers and marketplaces that need consistent product pictures at scale.

Internal links to related company profiles should be added only after the destination paths are confirmed live in the publisher’s own website. None are invented here.