CtrlK
Photoroom is an AI-powered photo editing platform headquartered in Paris, France, specializing in background removal, product photography, and generative image editing. Founded in 2019 by Matthieu Rouif and Eliot Andres, the company graduated from Y Combinator's Summer 2020 batch and has since grown into what it describes as the world's number-one AI photo editing app. The platform is available on iOS, Android, and the web, and also offers an API for enterprise integrations. As of 2024, Photoroom had surpassed 200 million downloads, processed over 5 billion images per year, and reached approximately $94 million in annual recurring revenue (ARR). The company raised $64 million in total funding across its seed, Series A, and Series B rounds, achieving a valuation of $500 million in February 2024.
Photoroom was co-founded in 2019 by Matthieu Rouif and Eliot Andres in Paris, France. The two met after Rouif, who had spent over a decade working on photo applications, took a machine learning course and began exploring ways to apply AI to image editing.
Rouif is a graduate of Ecole Polytechnique and Stanford University, where he attended one of the first iOS development classes in 2008 to 2009. Before Photoroom, he served in the French Air Force and then worked in product management at Stupeflix, a video creation startup that was acquired by GoPro. At GoPro, Rouif led image and video editing products. The founding motivation came from his frustration with tedious manual photo editing workflows. During his time at GoPro, he spent hours using the lasso tool to trace image edges when his team needed urgent marketing visuals.
Andres brought expertise in computer vision and deep learning. He is a published author on machine learning topics and an experienced mobile application developer. His technical background provided the foundation for building Photoroom's AI-driven editing capabilities.
The pair built the first version of Photoroom with a focused premise: using deep learning to automate the non-creative parts of photo editing, particularly background removal, so that anyone could create studio-quality product photos on a smartphone. They applied to Y Combinator and were accepted into the Summer 2020 batch, which was the accelerator's first fully remote cohort due to the COVID-19 pandemic.
Photoroom has raised a total of $64 million across multiple funding rounds, growing from a two-person startup to a company valued at $500 million.
Following its graduation from Y Combinator's Summer 2020 batch, Photoroom raised a seed round of $5.6 million. The round was led by Nicolas Wittenborn's Adjacent fund and included participation from Liquid2 Ventures. A notable group of angel investors contributed, several of whom brought significant AI and mobile subscription expertise.
| Investor | Background |
|---|---|
| Yann LeCun | Chief AI Scientist at Meta (Facebook) |
| Zehan Wang | Head of Twitter Machine Learning Cortex; co-founder of Magic Poney |
| Nicolas Pinto | Founder of Perceptio (acquired by Apple) |
| Holger Seim | Co-founder of Blinkist |
| Jacob Eiting | Co-founder of RevenueCat |
| John Bonten | Advisor for Calm and Spotify |
| Eric Setton | Co-founder of Tango |
The seed capital allowed Photoroom to accelerate product development and begin expanding its user base. The company demonstrated remarkable capital efficiency, scaling to $20 million in ARR on just $2 million of invested capital.
Photoroom raised $19 million in its Series A round, led by Balderton Capital. Angel investors from Facebook, Hugging Face, and Disney+ also participated, along with existing investor Adjacent. At the time of the announcement, Photoroom had reached 40 million app downloads across iOS and Android. The funds were earmarked for further software development, expanding generative AI features, and hiring top engineering talent in Europe.
In February 2024, Photoroom closed a $43 million Series B round at a $500 million valuation. The round was co-led by returning investor Balderton Capital and new investor Aglae Ventures, with Y Combinator also participating. By the time of this raise, Photoroom had been downloaded more than 150 million times and was processing over 5 billion images per year across more than 180 countries. The company announced plans to use the funding to invest in GPU infrastructure, secure imagery from leading image providers and photographers, and double the team size by the end of 2024.
The Series B announcement coincided with the launch of Photoroom's first custom foundation model, Photoroom Instant Diffusion.
Photoroom's technology stack combines multiple deep learning approaches, including image segmentation for background removal, diffusion models for generative image creation, and purpose-built models for tasks such as shadow generation, relighting, and upscaling.
Photoroom's background removal engine is the platform's core technology. Upon uploading an image, the AI detects and isolates the main subject (whether a product, person, or object) from its background. The system uses pixel-level image segmentation to classify each pixel as foreground or background. The technology preserves image resolution, sharpness, and edge precision so that the extracted subject remains crisp and natural.
A 2023 evaluation by Velebit AI found that Photoroom's API scored 70.8% accuracy on a standardized background removal benchmark, significantly outperforming Remove.bg, which scored 41.7% on the same test. Photoroom supports output in PNG (with transparency), JPEG, and WEBP formats.
Announced alongside the Series B funding in February 2024, Photoroom Instant Diffusion is the company's first custom foundation model, built specifically for product photography. Unlike many companies that fine-tune existing open-source models, Photoroom trained this model from scratch, making it one of the few companies in the world to have done so for this specific use case.
The model uses a Transformer-based architecture operating in latent space, similar in concept to DiT (Diffusion Transformer). The team chose this approach because Transformer architectures align well with GPU hardware and match the bandwidth-per-compute constraints needed for real-time inference. The model was designed from the beginning for multimodal capabilities rather than class-only conditioning.
Key technical details of the Photoroom ID model include:
| Aspect | Detail |
|---|---|
| Architecture | Transformer-based, operating in latent space (similar to DiT) |
| Parameter Count | Approximately 1 billion |
| Inference Latency | Sub-second (critical for mobile applications) |
| Training Dataset | Approximately 90 million images filtered from 1 billion candidates |
| Training Infrastructure | Mosaic Composer with PyTorch 2.x |
| Peak Training Throughput | Over 10,000 images per second |
| GPU Utilization | Around 80% peak consumption |
| Distillation | Uses LCM (Latent Consistency Model) techniques for speed |
The training approach incorporated image understanding and autocompletion via a masked prediction task, similar to BERT and MAE (Masked Autoencoder) frameworks. This allows the model to build precise knowledge of how image parts are related or can be altered. The team re-captioned training images using CogVLM and LLaVA rather than relying on CLIP-based alignment. NSFW content was filtered, and an in-house graphics detection system was used to curate the final training set.
The resulting model increased image generation speed by 40% compared to previous approaches while maintaining quality that internal testers described as a dramatic improvement.
Photoroom has contributed to the open-source AI community by releasing PRX (Photoroom eXperimental), a text-to-image diffusion model available under an Apache 2.0 license on Hugging Face. PRX is a custom, efficient MMDiT-like variant with 1.3 billion parameters for the flagship 1024-pixel model.
The 1024-pixel PRX model was trained in under 10 days on 32 H200 GPUs. It uses REPA (representation alignment) with DINOv2 features, the Flux VAE, and a T5-Gemma text embedder. Post-training methods include LADD (Latent Adversarial Diffusion Distillation), DPO, supervised fine-tuning, and Pref-GRPO. Rather than simply releasing model weights, Photoroom committed to documenting the complete training process through a research blog series, covering architecture design, training acceleration, and scaling experiments.
Photoroom offers a broad suite of AI-powered tools designed primarily for e-commerce sellers, marketers, and small business owners.
The flagship feature removes backgrounds from photos instantly. Users can keep the background transparent, replace it with a solid color (white, black, or custom), or swap it with an AI-generated scene. The tool handles complex subjects including transparent products like glass bottles, which traditionally pose challenges for segmentation models.
Users can generate realistic AI backgrounds in under one second. The system creates contextual scenes for product images, such as placing a coffee mug on a marble countertop or a pair of sneakers on a city sidewalk, without requiring any physical photography setup.
The Product Staging tool creates realistic product scenes using AI. It places products in contextually appropriate environments, simulating professional photography setups. This feature is aimed at e-commerce sellers who need lifestyle-style product images but lack the budget for professional photo shoots.
Photoroom's Virtual Model feature generates realistic images of products worn or displayed on AI-generated human models. This is particularly useful for fashion and apparel businesses. As of 2026, the feature supports customization of fit, lighting, poses, and backgrounds, and can display up to four products in a single generation.
This tool converts standard product images (including those taken with a smartphone) into studio-grade professional shots. The AI adjusts lighting, shadows, color balance, and composition to produce results that resemble professional photography.
Magic Retouch allows users to remove unwanted people, objects, or blemishes from images with a single swipe. The AI fills the removed areas with contextually appropriate content, similar to inpainting technology found in professional editing tools.
The expand feature extends the boundaries of an image by generating new content beyond the original frame, useful for adapting images to different aspect ratios or social media formats.
Photoroom supports bulk editing of thousands of images simultaneously through its Batch mode. This feature is critical for e-commerce sellers who need to process entire product catalogs with consistent styling. A November 2025 update delivered a 20x speed improvement for batch editing on the web platform.
Launched in November 2025, the AI Video Generator turns product images into short animated videos on iOS and Android. By January 2026, the feature was expanded to support text prompts that describe desired motion or visual effects.
The Brand Kit feature allows teams to share designs, colors, logos, and templates across an organization, ensuring consistent visual branding at scale.
An automated tool for apparel photography that removes mannequins from clothing images while preserving the garment's shape, creating the illusion that the clothing is being worn by an invisible figure.
Photoroom is available across multiple platforms:
| Platform | Details |
|---|---|
| iOS | Requires iOS 18.0+; works on iPhone, iPad, Mac (Apple Silicon), and Apple Vision Pro |
| Android | Available on Google Play |
| Web | Browser-based editor at photoroom.com |
| API | REST API for programmatic access and enterprise integrations |
The iOS app has a 4.8 out of 5 star rating based on over 217,000 reviews on the Apple App Store. The app file size is approximately 288 MB.
Photoroom offers a developer API that processes over 3 million images daily. The API provides two main tiers:
| API Tier | Capabilities |
|---|---|
| Remove Background API (Basic) | Background removal as a standalone service |
| Image Editing API (Plus) | Comprehensive editing including resizing, background replacement, lighting adjustment, shadow addition, product beautification, virtual model, ghost mannequin, flat lay generation, and image-to-video conversion |
The API uses standard HTTP calls and integrates with existing enterprise workflows, including PIM/DAM systems, CMS platforms, and e-commerce solutions. It is SOC 2 Type 2 certified for enterprise customers. A sandbox mode with watermarked results is available for free testing, along with 10 complimentary API calls upon account creation.
Notable API customers include Smartly, Printify, Faire, Bulgari, and Netflix.
Photoroom operates on a freemium model with multiple tiers:
| Plan | Target User | Key Features |
|---|---|---|
| Free | Casual users | 250 exports/month for Background Remover, Retouch, and Templates; limited AI feature access |
| Pro | Resellers and solopreneurs | Advanced AI tools (Product Staging, Virtual Model, Ghost Mannequin); 10x AI credit allowance vs. Free; 500 Batch exports/month; 1,000+ templates; high-resolution exports |
| Max | Small brands | Everything in Pro plus access to better AI models; 3x Pro's AI credit allowance; 1,500 Batch exports/month; Shopify listing creation and publishing; faster processing; priority support |
| Ultra (x1 through x10) | High-volume sellers | Everything in Max plus access to best AI models; Batch exports scale from 5,000 to 50,000/month; fastest processing; advanced AI workflows |
| Enterprise | Large organizations (200K+ images/year) | Custom plans; dedicated onboarding and customer success; SOC 2 Type 2 certified API; flexible credits with rollover; early access to new features |
Mobile subscription prices range from $5.99 to $249.99. Yearly billing offers approximately 47% savings compared to monthly billing. All plans include unlimited manual edits and single exports.
One of Photoroom's most visible public moments came through its collaboration with Warner Bros. and their marketing agency Bond on the official Barbie Selfie Generator, launched on April 3, 2023, ahead of the Barbie movie's theatrical release.
The selfie generator, hosted at barbieselfie.ai, used Photoroom's API to remove backgrounds from user-uploaded photos and replace them with imagery replicating the film's promotional posters. The tool allowed fans to create personalized Barbie-themed portraits. Warner Bros. tested 10 different background removal services and chose Photoroom; implementation of the API took less than one hour.
The campaign proved enormously successful, with the selfie generator used over 13 million times. Celebrities including Rihanna and actor Pedro Pascal publicly shared their Barbie selfies, amplifying the campaign's viral reach. The project demonstrated the practical value of Photoroom's API for large-scale consumer-facing applications beyond e-commerce.
Photoroom has demonstrated rapid growth across its key business metrics since its founding.
| Year | ARR | Growth |
|---|---|---|
| 2020 (August) | $1 million | First milestone |
| 2020 (December) | $2 million | 100% from August |
| 2022 (End of year) | $20 million | Scaled on $2M invested capital |
| 2023 (End of year) | $50 million | 150% year-over-year |
| 2024 (End of year) | $94 million | 89% year-over-year |
The company achieved profitability and maintains a customer payback period of approximately one month. Pro subscriptions account for approximately 70% of total revenue. Gross margins stood at roughly 85% as of 2023.
| Milestone | Date |
|---|---|
| 40 million downloads | November 2022 (Series A) |
| 100 million downloads | December 2023 |
| 150 million downloads | February 2024 (Series B) |
| 200 million downloads | Late 2024 |
| ~300 million users reached | 2025 |
Monthly active users grew from approximately 3.13 million in January 2023 to 12.89 million in October 2023, representing roughly 310% growth during that period. Photoroom reports a 75% user retention rate.
On the Apple App Store, Photoroom has ranked among the top three graphics and design apps in the United States, and at times held the number-one position. On Android, it has been the number-one photography app in the U.S. market. The platform operates in over 180 countries.
Major enterprise customers and partners include Shopify, Netflix, Lionsgate, Audi, Wolt, Zomato, Smartly, Printify, Faire, and Bulgari.
In May 2025, Photoroom completed its first acquisition, purchasing GenerateBanners, a startup specializing in automated image generation with precision text layout. The acquisition price was not disclosed.
The deal addressed a gap in Photoroom's API capabilities. While the API already powered millions of product images, enterprise customers needed template-based text composition to convert those product images into ready-to-run advertisements. GenerateBanners provided this functionality. Thibaut Patel, founder of GenerateBanners, joined Photoroom as Head of API.
The acquisition led to the launch of Visual Ads Automation, which Photoroom described as the first end-to-end generative AI solution covering both product imagery and automated text composition at catalog scale. Matthieu Rouif announced the acquisition from the Google I/O conference.
Photoroom has pursued several initiatives beyond its core product.
The company tracks CO2 emissions at inference time for its AI models and has published technical analysis detailing its measurement methodology. Photoroom positions sustainable AI as a "win-win-win" approach: greener infrastructure reduces costs, faster inference improves user experience, and both contribute to smarter model design.
Beyond the PRX text-to-image model, Photoroom has open-sourced a custom background removal model. The company has also published annual diversity reports and emphasizes transparency in its research process.
Photoroom maintains a "No DM" policy on Slack, meaning all team communications happen in public channels as part of a philosophy the company calls "radical openness." The team works in English even at the Paris headquarters. The company recruits globally and has published details about its approach to attracting and nurturing international talent.
Photoroom operates in a competitive market for AI-powered image editing, with several notable competitors.
| Competitor | Primary Focus | Key Differentiator |
|---|---|---|
| Remove.bg | Background removal | Single-purpose tool focused on one-off background removal tasks |
| Canva AI | All-in-one design platform | Extensive template library and integrations; background removal requires paid plan |
| Adobe Firefly | Creative AI integrated into Adobe suite | Deep integration with Photoshop, Illustrator, and other Adobe products |
| Clipdrop | AI image editing (owned by Jasper) | Suite of tools including relight, upscale, and cleanup |
| Claid.ai | E-commerce image automation | API-first approach for automated product image enhancement |
Photoroom differentiates itself through its combination of mobile-first design, e-commerce specialization, batch processing capabilities, and speed. Its freemium model with generous free-tier allowances (250 images per month for background removal) provides a lower barrier to entry than competitors such as Canva, which requires a paid plan for background removal. The company's focus on building proprietary foundation models for product photography, rather than relying solely on general-purpose models, also sets it apart technically.
| Detail | Information |
|---|---|
| Founded | 2019 |
| Founders | Matthieu Rouif (CEO), Eliot Andres (CTO) |
| Headquarters | 229 Rue Saint-Honore, Paris, 75001, France |
| Employees | Approximately 100 (as of 2024) |
| Total Funding | $64 million |
| Valuation | $500 million (February 2024) |
| Y Combinator Batch | Summer 2020 (S20) |
| Key Investors | Balderton Capital, Aglae Ventures, Y Combinator, Adjacent, Liquid2 Ventures, Kima Ventures, FJ Labs |