Doji Raises $14M for AI-Powered Virtual Try-On Technology

The Rise of AI-Powered Virtual Try-Ons

For a considerable period, major technology corporations, including Amazon and Google, have been dedicated to enhancing the engagement of virtual try-on experiences. Their aim is to stimulate increased online shopping activity among consumers.

Numerous startups have also focused on addressing challenges related to fashion inspiration and accurate fit using artificial intelligence technologies.

Introducing Doji: A New Approach

A recently established startup, Doji, is now entering this competitive landscape with an application specifically designed to make apparel try-ons both enjoyable and socially interactive.

The app functions by generating a personalized avatar for each user. Subsequently, it presents a variety of curated looks intended to inspire new clothing purchases.

Initial feedback from early users has been notably positive.

Securing Seed Funding for Growth

Just days after its public launch on the App Store, Doji has announced the successful completion of a $14 million seed funding round.

This funding round was spearheaded by Thrive Capital, with additional participation from Seven Seven Six.

The company intends to allocate these funds towards the refinement of its AI models, according to official statements.

The Role of AI Advancements

The development of an application like Doji is directly attributable to the recent significant advancements in the field of artificial intelligence.

Doji leverages its proprietary diffusion models to create highly personalized avatars and to deliver more realistic clothing try-on simulations.

The Founders and Their Vision

The startup was founded last year by Dorian Dargan and Jim Winkens, a collaboration reflected in the company’s name, Doji (a combination of Dorian and Jim).

Dargan brings prior experience from Apple, where he contributed to the development of VisionOS, and from Meta, where he worked on games and experiences for the Oculus Quest platform.

Winkens previously served as a researcher at DeepMind and also contributed to a consumer product based on generative AI at Google.

The two founders initially connected on Twitter/X around 2022. A shared passion for the fashion industry led them to explore collaborative side projects.

The release of the Lensa avatar creation app proved to be a pivotal moment, as Dargan observed the strong emotional connection users developed with their avatars.

“Lensa focused on stylized avatars with diverse themes,” Dargan explained to TechCrunch. “We questioned whether we could replicate this for fashion, but with a focus on photorealistic results.”

Industry Perspective on Doji

Miles Grimshaw, a partner at Thrive Capital, expressed his enthusiasm for Doji’s application from the moment he first experienced it.

He highlighted the potential of diffusion models to create a “virtual mirror” for trying on clothes.

“The current online shopping experience, requiring navigation through numerous links, can be cumbersome. Doji presents an opportunity to transform shopping into an enjoyable experience by placing the user at its core.”

Grimshaw also emphasized the app’s social dimension, noting its potential to encourage users to share different looks with their social circles.

Experiencing the Doji Application

Currently operating in an invite-only capacity, Doji guides its users through a process involving the capture of six selfies and the upload of two full-body photographs. These are utilized to generate a personalized avatar. The avatar creation process typically requires approximately 30 minutes, after which users receive a notification indicating its completion.

During the initial setup, users are also presented with the opportunity to select their preferred brands. This customization allows the app to prioritize displaying items from those brands within the user’s feed.

Several other applications, including Glance – a lock screen app supported by Google – have attempted to deliver AI-driven shopping suggestions based on a single image. However, Doji has chosen a more comprehensive onboarding procedure. This approach aims to produce higher-fidelity avatars and facilitate more accurate clothing recommendations.

The app, by default, presents users with clothing suggestions tailored to their avatar. Users can browse through a variety of tops and bottoms, and by selecting an item, they can visualize it on their avatar.

Furthermore, the platform allows users to submit links to apparel found elsewhere on the web, enabling them to assess its suitability for their virtual representation.

During testing, it was observed that the avatar’s depiction sometimes presented a slimmer or taller physique than the user’s actual proportions. (Doji acknowledges that users have the ability to refine their avatar through a new image set if the initial results are unsatisfactory.)

While the application facilitates the visualization of different clothing items, it currently lacks the capability to accurately predict garment fit. Addressing this limitation is a future development goal for the company.

The development team is also focused on accelerating the virtual try-on process and streamlining the purchasing experience. The intention is to integrate direct purchasing within the app, rather than redirecting users to external websites.

Available in over 80 countries, the startup has not yet announced a timeline for removing the invite-only restriction and opening the app to the general public.

Dargan emphasized that Doji distinguishes itself from competing try-on applications by offering an enjoyable experience for clothing exploration. He also expressed confidence in Doji’s superior human representation, attributing this to his extensive experience in avatar design.

“Previous attempts to realize this concept have often been hindered by either a lack of sufficient technological expertise or a deficiency in aesthetic sensibility regarding what truly enhances a person’s self-perception through imagery,” Dargan stated. “We have made substantial investments in foundational technology to ensure that your self-image within the app is both captivating and motivating.”