VerSe bets on SparkStation to make AI video cheaper and more consistent

A 60-second brand film that costs Rs 10 lakh to Rs 25 lakh to produce could be made for Rs 50,000 to Rs 75,000 and delivered in about 24 hours...if VerSe Innovation’s latest AI bet lives up to its promise.

The parent company of local-language content aggregator Dailyhunt and short-video platform Josh on Saturday (August 29) launched SparkStation, an AI-powered content operating system that puts ideation, scripting, casting, storyboarding, music, editing, rendering and distribution into one workflow.

VerSe has committed more than $30 million over two years to GPU infrastructure, AI orchestration and specialist talent. Its larger aim is to make production capabilities that required a studio, crew and expensive post-production available to brands, filmmakers and creators.


Also Read: Hollywood, Bollywood and AI: Who controls storytelling now?

VerSe co-founder and CEO Umang Bedi said in a conversation with ET Online that the company sees production costs as one of the biggest constraints on what gets made.

“Production cost has been the invisible ceiling on what stories get made. There are so many stories and ideas that get shelved primarily because of the lack of budgets,” Bedi told ET Online following the launch of the platform.

How the cost comes down

There is no fixed price for an AI-generated film on SparkStation. In a demo to ET Online, its budget estimator put a one-minute commercial at about $200 (roughly Rs 20,000). Bedi said a higher-quality output could cost Rs 25,000 a minute, against a potential traditional cost of Rs 2 lakh.

The bill changes with the camera, lens, movement, AI model and production complexity selected. In the demonstration, the estimated token requirement ranged from 16 to 406 as these choices changed.

“The cost of the same task can change depending on the camera, lens and AI model,” Bedi said.

SparkStation acts as an orchestration layer, sending tasks such as creating a face, camera movement, voice or language to different AI models. Creators will pay through credits, studios and brands through seat-based subscriptions and larger customers through enterprise licences and API access.

Solving AI video’s continuity problem

Generating one convincing AI video shot is no longer the only challenge. The harder part is producing the next shot with the same person, clothes, location and objects.

VerSe says SparkStation addresses this by locking five elements across scenes, including face, expression, wardrobe, location and props. Bedi called this the company’s “secret sauce”.

“That is the IP we’ve really solved for,” he said.

The platform supports more than 60 languages and, according to VerSe, can adapt lip movements, emotions and local idioms rather than simply translate words. Bedi said the company has accumulated text, image, audio and video data through its platforms and more than 10,000 publishers. It has primarily trained its own models on content data in 16 languages, he added.

SparkStation is also being positioned as more than another AI video-generation tool. Projects created on the platform can be exported into professional editing software such as Premiere, Final Cut and DaVinci Resolve.

Also Read: A once wary Hollywood slowly warms to AI

SparkStation has been used within VerSe for more than six months and is already generating revenue, according to the company. Bedi said it has around 15 clients, including two filmmakers and two production houses. VerSe separately said the platform has worked with more than 40 brands and over 20 social-media influencers.

The company did not name the clients as these are live production engagements under confidentiality. However, it did showcase work created on the platform by brands such as personal care brand Dhathri, OTT platform Zee5, etc.

Beta access began on August 29, with general availability planned for October 1 and an enterprise API expected on November 1. VerSe aims to serve more than 100 studios, 1,000 brands and thousands of creators globally.

Risk of more synthetic content

Lower costs could become a double-edged sword. As platforms try to limit AI slop, SparkStation’s larger test may be whether people want to watch what it helps make, not just how cheaply it makes it.

Bedi acknowledged that trust is becoming increasingly important as AI-generated advertising and social content become harder to distinguish from traditionally produced material.

India’s amended IT Rules require synthetic content to carry prominent labels and, where technically feasible, permanent metadata or other provenance mechanisms.

Bedi said VerSe is building licensing tools that would allow people to license their voices and likenesses and potentially earn royalties when they are used in AI-generated content. “Copyright is a huge issue,” he said.

VerSe said SparkStation adds a backend watermark using SynthID. It also said non-enterprise customer data is currently stored in India, while enterprise customers using the upcoming API can store data in their own environment. Bedi added that the platform complies with India’s Digital Personal Data Protection (DPDP) framework.