Generative AI models have achieved a basic ability to produce at least a passable video from a single image or short sentence. Companies building products around these models are claiming that anyone can make a great promotional video if they have some images or footage – and videos usually perform better than static images or documents.
Tiger Global-backed Peak XV and Avataar released a new tool called Velocity on Monday. It creates product videos directly based on a product link. The company would go up against the likes of Amazon and Google, which are also experimenting with AI-powered video tools for advertising.
The startup ethos behind these tools is similar to video creation tools like Lica or Sythesia. Creating videos is expensive, so companies can’t make videos for everything. However, thanks to artificial intelligence, the cost of creating videos is getting lower and you can generate clips in bulk.
Avataar was founded by former Microsoft and Deutsche Bank employee Sravanth Aluru in 2015. The company, which has raised over $55 million in funding, previously focused on creating interactive experiences, AI-powered imagery and 3D models for electronic commerce. The startup is now focusing more on using AI to create videos for brands.
The company said 3D models and experiences have a higher conversion rate than videos. However, many more people engage with video, so even with lower conversion rates, videos win because of scale.
In February 2024, the company began commercially testing its AI-powered video creation tool with customers. But at the time, Avataar still had to involve people in the process to help these brands create videos. The company began testing its fully automated Velocity tool in September 2024.
His tool is already being used by HP, Victoria’s Secret, Lowe’s, Newegg, Marina, TVS and Bajaj to generate various product videos.
Avataar said that as brands use its services and tools to create custom videos for their core offerings, Velocity will be useful in creating product clips for products that aren’t strategic enough to get a budget to create. videos. The startup wants to drive better conversion rates for these products through AI-generated informational videos.
“We can cover entire catalogs and create videos for all the items in them through Velocity. This would improve the customer experience and also drive better conversion through movement and storytelling around the product,” Aluru told TechCrunch.
The company also has an API that companies can integrate with their platforms to create product videos for automatic listings.
Aluru said Avataar’s models or solutions differ from others because they have collected a lot of data about products through 3D rendering. He noted that brands can’t afford to have glitches in their product launch videos, and the company has made sure of that.
“Our models understand different product sub-categories and their associated attributes. Our tool will highlight different attributes of a hammer compared to a smartphone. We focus heavily on better storytelling using video as a medium with an ultimate goal of driving purchases,” he said.
Through its videos, Avataar also tries to create aspirational value for the products apart from providing details to the customer.
The company noted that it has built in guardrails around brand safety and product inaccuracies, so a model can check for inaccuracies in a creation and reject the product created by the tool.
Amir Konigsberg, an investor and founder of the aborted e-commerce startup Twiggle, said there is an opportunity to make videos for sale online, but they need to be unique enough to stand out.
“Generating videos from product links is a natural extension of eCommerce pipeline automation. Video is a proven driver of engagement and conversions, and scaling this efficiently is smart. However, there are already several companies dealing with aspects of this space – such as generating backgrounds, after effects, renderings of products from different angles, cutouts or photorealistic shadows. The challenge here lies in differentiating, ensuring consistency and respecting licensing and attribution standards,” he told TechCrunch via email.
Konigsberg added that brands didn’t adopt AI-generated video solutions until now because it was challenging to maintain quality at scale. However, he hinted that due to recent advances in models, these tools could become effective for mass use.
