Dresma AI generates high-performing, copyright-compliant product visuals for large retailers at scale with Google Cloud and generative media models.
Dresma AI generates high-performing, copyright-compliant product visuals for large retailers at scale with Google Cloud and generative media models.
A world of products in pictures and videos
Google Cloud provides an integrated platform with state-of-the-art LLMs and APIs backed by easy-to-execute workflow orchestration and scalable compute.
Nishka Sinha
Co-founder and CMO, Dresma
How many ecommerce websites did you visit today? And how many pages did you scroll through before you made a purchase? An estimated 600 million items are available from Amazon alone — each one competing for your attention with product photos and videos. That's a heck of a lot of content.
"Even large brands struggle to keep up with the demand for product-related visuals," states Nishka Sinha, the co-founder and CMO of AI startup Dresma. "Our company's mission is to help brands leverage artificial intelligence to create all the visual content they need to convert browsers to buyers."
Dresma's platform tracks performance metrics such as impressions, engagement, reviews, and sales on platforms like Instagram, Pinterest, and Amazon, where their customers market and sell their products. By marrying this data to the product videos and photography in use, the platform determines what type of content will perform best.
That's when Dresma's AI tools go to work. By amalgamating content from a customer's digital asset management (DAM) system, its brand guidelines, the requirements of each website or marketing channel, and its own platform's aggregated data, Dresma AI creates high-performing still photos and videos at scale.
And for large retailers with thousands or even tens of thousands of SKUs as well as markets across the globe, the scale can be enormous, the variety vast, and the level of detail daunting.
"In product photography, there's no room for error, but the large language models we'd been using had trouble accurately reproducing attributes like textures and colors," Sinha acknowledges. Dresma also had concerns regarding the provenance of the training data used for those LLMs — an especially sensitive issue when synthesizing faces for fashion photography.
And its previous cloud provider couldn't keep up with the demands of a platform that processes 20,000 to 30,000 images and videos daily: The data-transfer bottlenecks and high failure rate on API calls to customers' DAMs limited content production and impeded Dresma's growth. "We needed an integrated platform with state-of-the-art LLMs and APIs backed by easy-to-execute workflow orchestration and scalable compute," she recounts. So Dresma turned to Google.
Precision AI content generation and fast, reliable throughput
Joining the six-week Google Startup Accelerator Program was a turning point for Dresma. "Among the people we worked with was an expert in generative AI video and photography who helped us tremendously with our platform's evolution and development," Sinha recalls.
Dresma now hosts its platform on Google Cloud, using its durable APIs to link its customers' DAMs directly to Gemini models. Reference images are parsed using Cloud Vision AI, and those insights — about everything from shape, color, and texture to lighting and backgrounds — are fed directly into Gemini-based LLMs. Dresma is also relying heavily on Veo for video creation and on Imagen for the creation of product images.
With Google Cloud AI tools, we can create more brand-consistent assets more accurately and efficiently, and we’re confident the content we create is copyright-compliant.
Nishka Sinha
Co-founder and CMO, Dresma
Google virtual machines scale compute up and down automatically with demand, increasing capacity and saving costs. Workflows for each customer and each content type are orchestrated using Cloud Run, which is also used for the real-time data feeds piped into the Dresma platform from various marketing and ecommerce sites.
The consolidation of its IT and AI operations on Google Cloud took less than a month; the results of moving to an integrated AI ecosystem have been dramatic. "Our overall LLM reliability and platform performance is exponentially better," Sinha reports.
The high success rate for API calls has increased throughput by a factor of five and decreased turnaround times for content generation 40%. Better image quality and fidelity means fewer do-overs: Content volume production is up 400%. Lower IT infrastructure and compute costs have also upped profit margins 15%.
"With Google Cloud AI tools, we can create more brand-consistent assets more accurately and efficiently, and we're confident the content we create is copyright-compliant," she explains. "As a result, our customers send us more work, and we've augmented our offerings."
Technology that attracts new customers
Many of our prospects use Google products themselves, and knowing Google Cloud forms the foundation of our platform and LLMs immediately instills trust in our company and makes it easier to close new business.
Nishka Sinha
Co-founder and CMO, Dresma
The Google Cloud holistic software development toolset has also accelerated product development and halved time-to-market for new platform features. Those features include a moodboard that visualizes the performance data Dresma collects, and a soon-to-be-released data platform for social media and direct-to-consumer marketplaces.
And Sinha discovered that migrating to Google Cloud came with an unexpected bonus. "Many of our prospects use Google products themselves, and knowing that Google Cloud forms the foundation of our platform and LLMs immediately instills trust in our company and makes it easier to close new business," she declares.
The philosophy behind Google AI also complements Dresma's. "It's not merely AI for AI's sake," Sinha observes. "Google uses AI to solve real-world problems, and Dresma does the same for our customers." That kinship extends to the company's work with their Google Cloud team, which ensured a smooth migration and continues to provide go-to-market, product development, and technical support.
"Other LLM companies kept us at arm's length, but at Google, there's always an expert eager to provide any technical support we need to meet our customers' needs," Sinha says, adding that the communication is a two-way street: "Google is also interested in how we're using their technologies and elicits input on ways it can make their products even better."
For a startup embarking on its Series A fundraising, that kind of support can make all the difference. "We're always pleasantly surprised by how engaged our Google team remains," Sinha concludes. "Knowing we're backed by a company that's genuinely invested in our success and helping us grow is a very, very reassuring feeling."
Dresma is an all-in-one AI content analysis, creation, and optimization platform for generating product photos and videos that convert browsers into buyers on any marketing channel or ecommerce site.
Industry: Retail
Location: India
Products: Cloud Run, Cloud Vision AI, Compute Engine instances, Imagen, Gemini, Google Cloud, Veo