Unique Research · Apps · January 8, 2025
Generative AI models can already turn a single image or a short sentence into at least a passable video. Companies building products around those models claim that with a picture or a recording, anyone can make a polished promo — and that video usually beats static images or documents.
Avataar, backed by Peak XV and Tiger Global, on Monday launched a new tool called Velocity that generates product videos directly from a product link. The move puts Avataar in competition with players such as Amazon and Google that are also testing AI-driven video tools for advertising.
The startup thesis behind these tools is similar to video creators such as Lica or Synthesia. Video is expensive, so companies cannot produce it for every SKU. AI is bringing costs down and making batch clips feasible.
Avataar was founded in 2015 by Sravanth Aluru, formerly of Microsoft and Deutsche Bank. It has raised more than $55 million and previously focused on interactive experiences, AI images, and 3D models for e-commerce. It is now more focused on AI video for brands.
The company says 3D models and experiences convert better, but more people watch video — so even at lower conversion, video wins on scale.
In February 2024 Avataar began commercial testing of its AI video tools with customers, still with humans in the loop to help brands produce clips. In September 2024 it began testing the fully automated Velocity tool.
Customers now using its tools for product video include HP, Victoria’s Secret, Lowe’s, Newegg, Marina, TVS, and Bajaj.
Avataar says brands still use its services for custom video on flagship products; Velocity is meant for SKUs that lack the strategic weight to justify a video budget. The startup wants information-rich AI video to lift conversion on those items.
“With Velocity we can cover the entire catalog and create video for every item. That improves customer experience and raises conversion through motion and storytelling around the product,” Aluru told TechCrunch.
The company also offers an API so other platforms can auto-create video for product listings.
Aluru said Avataar’s models stand out because of the product data collected through 3D rendering. Brands cannot tolerate glitches in product showcase video, and Avataar has worked to ensure that.
“Our models understand different product subcategories and their attributes. Versus a smartphone, our tool highlights different attributes for a hammer. We care a lot about using video as a medium to tell a better story — ultimately to drive purchase,” he said.
Beyond detailed information, Avataar also tries to create aspirational value for products through its videos.
It says it has built guardrails for brand safety and product accuracy so the model can check creations for inaccuracies and refuse bad outputs.
Amir Konigsberg, founder of shuttered e-commerce startup Twiggle and an investor, told TechCrunch by email that video for online sales is a real opportunity — but clips must be distinctive enough to stand out.
“Generating video from a product link is a natural extension of e-commerce automation. Video has proven to lift engagement and conversion, and scaling that efficiently is smart. But many companies already handle pieces of this space — backgrounds, post effects, multi-angle renders, cutouts, realistic shadows. The challenge is differentiation, consistency, and meeting licensing and attribution standards.”
Konigsberg added that brands have so far been slow to adopt AI video because quality is hard to keep at scale — though recent model progress may make these tools fit for large-scale use.
— END —