Manvitha Mallela and Malavika Reddy, co-founders of Flock AI, started their company because they didn’t see themselves represented in their online shopping journeys.
TikTok's Latest Tool Enables Brands to Use AI-Generated Avatars in Ads
The two women faced major pain points when trying to determine whether a fashion, apparel or beauty item would suit their body or skin tone. With those issues in mind, they decided to create a way to help brands use artificial intelligence to bring greater representation to product imagery—and to analyze how best to reach companies’ diverse consumer bases.
Now, Flock has raised $1.5 million in pre-seed funding to continue that mission. Outlander VC led the round, with participation from Red Swan Ventures, F4 Fund, Hustle Fund and Dorm Room Fund.
Mallela, who serves as the CEO, has a background in retail, having previously worked at Bloomingdale’s , Walmart and Jet.com . Meanwhile, Reddy, who serves as the CTO, has a technological background and had experience working as a Microsoft engineer. When the two joined forces—and identified a similar problem they had with shopping—they began building what felt like a natural solution.
“When I met [Reddy], she had just moved from India to the US for our MBA program, and she was like, ‘I just feel like I never see content that relates to me or represents me—how do you even shop online?'” Mallela told Sourcing Journal. “With her background in AI and computer vision …we really came together and were like, it seems like there’s something that we can uniquely do here around how can we solve the limitations of [the] physical photoshoot so we can infinitely scale your on-model imagery, and then solve the needs for customers.”
Mallela describes Flock as an end-to-end visual commerce platform. It leverages a combination of proprietary deep learning models and open-sourced generative AI tools to help brands create product imagery that suits a variety of moments, from product detail pages, to Meta ads and social media posts.
The resulting images, which the Flock team is able to turn around within 24 to 48 hours, display models wearing apparel items, accessories, footwear or beauty products.
With its new funding, the startup has plans to add more members to its teams in the U.S. and India, where Reddy lives part of the time. Though it has a technology-focused employee base in India, its primary markets remain the U.S. and the EU. To help support its growth, Mallela plans to add headcount to the company’s partnerships team, which could help bring in more clients.
Mallela said that the technology has applications far beyond retailers, because, to develop the rendering, Flock needs very little. Brands and retailers can submit lay-flat images, on-model images, 3D images or other digital assets. Most often, though, it sees lay-flat images, which brands often have on hand and aren’t overly expensive to produce.
“The [retailers] of the world, the resellers of the world, the rental companies, they’re never going to be able to create those 3D models, just because they don’t have pattern information and all of that like the brands do. That’s why we can uniquely work with these guys, too—because we don’t ask for any complex inputs,” she said.
And on the model side, the company has built a proprietary algorithm it calls AI for All to help bust up biases typical of many large language models and generative AI systems today.
“If you go to any other image creator and you try to get a model, you’re going to get someone that’s a certain skin tone, certain height, body shape, all of that. We’ve really taken that away,” Mallela told Sourcing Journal. “Our AI for All algorithm basically captures [varying] ages [and] skin tones, [as well as] things like wrinkles, acne, poor texture—stuff that makes us like uniquely imperfect. We’ve gotten rid of those inherent race and gender biases that exists in other image creators, and this is why we can do really cool things.”
Mallela said the pre-seed money will also help Flock purchase reliable, non-biased, third-party data to continue training its technology.
Byte-Sized AI: JD Sports Amps Search; Circularity Continues to Show Up in AI Systems
While it can generate completely fake models, Flock also has the ability to digitize real-life models and people, should brands choose to hire them for their digital likeness . For some brands, that seems to be a way to cut costs while avoiding the kind of backlash names like Levi’s have faced for using fully AI-generated models.
However, Mallela said, any use of a real person’s digital likeness must be cleared by that person in order for Flock to release the images to a company. That insistence on Flock’s part could help brands stay in keeping with the New York Fashion Workers Act , which includes provisions about models’ digital likenesses. The legislation has passed both the state’s governing bodies and awaits signature by Governor Kathy Hochul.
Reddy has developed a piece of the system that can identify if a generated image bears striking resemblance to a celebrity or model. So, for instance, if a brand wants to generate a redheaded model with hazel eyes, the system will automatically check that the output doesn’t resemble a real-life person with those characteristics, like Lindsay Lohan.
When Flock works with a client, it can serve up analytical recommendations on the type of images a company should look to create.
“We can give you preemptive recommendations like, ‘Oh, you’re creating a Meta ad for this product, maybe you should maybe localize for this area. You want X, Y and Z model characteristics, half-body shots, looking at the user, with this color background.'”
That, she said, helps with engagement and conversion for Meta and Google ad targeting. Already, brands the company works with have seen a 10x reduction in image-generation costs and over a 30 percent increase in conversion.
As the startup continues building relationships, it plans to use some of its pre-seed funding to enrich the technology based on companies’ needs. One such upgrade may include new types of backgrounds for images. At the moment, Flock’s images appear as if they have been taken in a studio. The startup has not yet mastered image generation for editorial or location-based shots, which Mallela said is because it wants to ensure the products displayed are done so in a hyper accurate way.
“It’s hard right now to be able to maintain the accuracy of the sleeve length, where the dress hits you, the buttonhole, all of that while placing models in a virtual environment, where they might be leaning on a tree or running through the fields,” she said. “At the moment, it’s not something that we focus on or offer, but it is something down the line the product development that we [would like] to be able to achieve.”
Mallela said she and Reddy faced quite a few “nos” from potential investors before finding those that believed in the idea. Data from Pitchbook shows that in 2022, women of color received just 0.39 percent of available VC funding.
Despite their early obstacles, Mallela said she knows she and Reddy are the ones that can bring the solution to market in a big way.
“We as POC shoppers needed to see this solution out there, too—and built in a way that’s ethical, built in a way that’s thought through. I think that’s something that really stood out compared to anybody else that might be playing or thinking about the space that really worked in our favor,” she said.
Salesforce: Platforms Like Shein and Temu Could Grab 21 Percent of Holiday Spending
Most Popular
Comments / 0