Alexandre de Vigan: Nfinite aims to become the global leader in synthetic image creation

A start-up founded in 2017, Nfinite has seen significant success with the launch of its SaaS e-merchandising platform, which allows for the automatic creation of visuals from 3D product models. Alexandre de Vigan, its founder, discusses his journey and ambitions and the imminent launch of a technology for creating synthetic data, poised to revolutionize the world of AI by offering its clients (GAFAM, app creators, retail giants) the ability to create customized, high-quality visuals in an unlimited, cost-effective, safe manner while avoiding copyright issues.

You founded Nfinite in 2017, a company specialized in creating imagery for the real estate sector, now focusing on providing 3D visuals for e-commerce businesses through CGI (computer-generated imagery) technology. How does CGI relate to generative AI? 

CGI involves generating computer images, whether it’s based on a 3D engine, Computer Vision tools, or artificial intelligence. It’s essential to distinguish between the creation technology and its result: the result being the image (CGI), which can be generated through a variety of combined or independent methods.

At Nfinite, our team of engineers has spent seven years developing and enhancing a platform for creating limitless, photorealistic, and controlled computer-generated images. Initially, seven years ago, this platform was based on 3D and computer vision. Today, we’ve integrated machine learning, artificial intelligence, and generative AI to expand our creative capabilities while ensuring a level of image quality and control that is unique on the market.

Nfinite’s journey has been one of successive trials and errors until the launch of our SAS platform for retailers in 2021. This marked a significant acceleration: a first $15 million Series A funding round, a tenfold increase in revenue, and expansion to the United States in the same year, followed by a second $100 million Series B funding round and a threefold revenue increase in 2022. What are your goals by 2026, considering the e-commerce sector’s projected 50% growth to $7.4 trillion? 

Nfinite aims to become the global leader in synthetic image creation (CGI). Since 2021, we’ve expanded from France to the United States and now across Europe in the retail and e-commerce verticals, enabling merchants to create unlimited images/videos to boost sales across all channels.

While we continue to grow strongly in e-commerce—a sector experiencing explosive growth in product experience consumption (visuals, videos, 3D, etc.)—we recognize our platform’s potential for various other applications: Advertising, Communication, Security, Technology, Industry, and notably, increasingly direct integration with artificial intelligence (either downstream, combining 3D and AI to serve enterprise clients, or upstream to enhance AI performance via 3D and synthetic images).

Do you plan to continue this organic growth strategy, or do you foresee potential partnerships in the coming years? 

Currently, we’re pursuing organic growth through three avenues: expanding geographically, entering new markets, and growing our client base in existing markets. However, we believe in strategic partnerships where complementary offerings provide unique value to existing or future clients.

Your clients primarily come from the furniture and decoration industry. What proportion of your revenue does this segment represent? In which other sectors are you present or planning to enter? 

Until the end of 2023, our clients were mainly large retailers in the US, France, and Europe. These clients expanded beyond furniture and decoration into various sectors: FMCG, Food & Beverage, Home, DIY, and fashion accessories. Since late 2023, we’ve observed significant acceleration in other segments (GAFAM and Big Tech) using our platform to generate synthetic training data. Currently, 50% of our clients are retailers, and 50% are AI players seeking training data from Nfinite to enhance their LLM performance or visual generation applications via generative AI.

Your company utilizes a database of AI-generated data to provide visuals and images to retailers. What are the advantages of this technology over traditional product photography methods? 

Nfinite is a visual creation automation platform. We enable our retail clients to automatically create their entire product experiences via our platform without any logistics or physical elements. Everything is virtual. This saves time, enhances efficiency, and reduces costs. Importantly, it improves performance by creating complete, consistent, and immersive product experiences at scale for entire product catalogs, significantly boosting e-commerce performance (traffic, engagement, cart additions, and sales).

What sets you apart from free competing solutions like Blend and PixelBin’s Generative Background Creator, which also leverages generative AI? Are your targets the same? 

Nfinite distinguishes itself by its ability to create photorealistic visuals at a massive scale using various technologies mentioned above. It caters particularly to large enterprise players with needs for tens of thousands of references. Our platform enables them to generate dozens of visuals per product while ensuring absolute consistency and brand integrity. This approach not only saves significant time and cost compared to traditional photoshoots but also allows infinite customization and variability, crucial for meeting consumer expectations in purchasing experiences.

Simultaneously, you’re investing in technology capable of generating synthetic data, specifically entirely artificial images. Could you elaborate on what synthetic data is? How does this technology differ from DALL-E, for instance, OpenAI’s artificial image generator (ChatGPT, Sora…)? What are its advantages? 

Synthetic data, especially in the form of 100% artificial images, are computer-generated information mimicking real data, used to train generative AI models like OpenAI’s DALL-E. These and many other players are in a race to build the best visual generation models using neural networks. To ensure the commercial success of these models, they need to train them with a phenomenal amount of data, including images meeting very specific criteria: quantity, quality, lack of intellectual property, specificity, etc. By generating sets of synthetic images “on demand,” Nfinite enables these players to enrich their models without the constraints of traditional image sets they typically have access to (search engines, private databases, etc.).

Gartner analysts predict that 60% of data used to train AI systems worldwide will be synthetic by the end of this year, compared to 1% in 2021. How do you plan to establish yourself in this exploding market in the coming months? 

Scientific literature (white papers, lab studies, and AI company communications) suggests that real data will be insufficient in the short term to train AI engines, for two main reasons: the next iteration of AI engines will require much larger training sets than current engines, and applying AI solutions to business needs requires specific training on data tailored to the company’s use cases. Synthetic data is the only solution to these needs. Consequently, the time to market is perfect for players positioned in synthetic data creation, addressing the following equation:

• Volume
• Quality
• Specificity/Controllability
• IP
• Efficiency
• Tagging

What types of clients are you targeting? Tech giants, app creators, retailers? Do they all understand what synthetic data is and how indispensable it will be in the coming months and years? 

We have three potential client types: visual LLM publishers (Big Tech) needing large quantities of training data for foundational models, visual AI application publishers needing training data for vertically training their applications for specific use cases, and companies deploying/implementing AI applications to address their operational needs (supply chain, security, marketing, etc.). Each use case requires training with data specific to the case being addressed. This is what we call the “last mile quality gap,” bridging the gap between the LLM and/or the application and the company’s specific needs.

These three client typologies differ in their understanding of AI challenges, particularly synthetic data—those most “advanced” being, naturally, LLM publishers.

Have you read?
Navruz Avloni: How CEOs Can Cultivate Trust and Safety in the Workplace.
Best CEO Forum Companion Hack? Visibility.
Best Books for Examining Modern Motherhood.
Luxury Living for American Expats: Top Dubai Neighbourhoods and Lifestyle Amenities.
Accept Result and Strategize.