SUMMARY
Trained on high-quality synthetic data with a diverse set of identities, Kroop AI’s deepfake detector tools can accurately differentiate between real and fake audio and video
The startup works with clients in the ecommerce and pharma industries, but its tools have found the biggest use case in the BFSI sector
To date, Kroop AI has secured $230K in funding from investors such as 100X.VC and LetsVenture, along with support from several angels and family offices
Last year, an AI-generated deepfake video featuring Indian actor Rashmika Mandanna went viral, sparking widespread concern about the ethical implications of emerging technologies like AI and open-source GenAI models, as well as the regulations governing their use. Following Mandanna’s incident, several other celebrities became victims of deepfake manipulation.
Concerned with the rise in the number of such cases, Prime Minister Narendra Modi called for global regulation of AI to address the issue.
Interestingly, the founder and CEO of Ahmedabad-based Kroop AI, Jyoti Joshi, saw this coming a few years ago and, therefore, started to work on a deepfake detection solution in anticipation.
Founded in early 2021, Kroop AI is one of the first few GenAI startups in India that is on a mission to assuage this world from the deepfake menace. Kroop AI’s tools enable users to detect whether an image, video, or voice is AI-fabricated, AI-generated or real.
Trained on high-quality synthetic data with a diverse set of identities, Kroop AI’s deepfake detector can accurately differentiate between real and fake audio and video.
For the uninitiated, deepfake is synthetic media, generated using deep learning techniques to manipulate or create video, audio, or images to present something that has not happened.
The technology behind deepfakes primarily relies on deep learning algorithms, typically Generative Adversarial Networks (GAN). It engages two AI algorithms against each other with the generator network creating a manipulated image and the discriminator trying to detect whether the generated image is fake. This process continues until the discriminator can no longer differentiate between the real and the fabricated images.
It goes without saying that with the boom of GenAI platforms that can create video, audio, and images at the cues of a few prompts, the risk of deepfake has increased, and players like Kroop AI have a major role to play.
But Kroop AI Is More Than Just A Deepfake Detector
Kroop AI’s genesis dates back to the Covid-19 era when the GenAI wave had yet to hit Indians. This was the time, around the US elections in 2020, when instances of the misuse of deepfake technology were frequently making news.
Driven by the realisation of the significant consequences that could arise from the misuse of this technology, Joshi, also an AI scientist, seized upon the idea of developing a deepfake detection solution.
Recognising the sparse presence of players in the deepfake detection market, Joshi saw an opportune moment to enter the space. However, at the time, awareness of the risks associated with deepfake technology in India was limited, partly due to the delayed adoption of AI technology and GenAI in the country.
As Kroop AI began assembling vast datasets to train its deepfake detection models, the startup also recognised a significant opportunity in the area of ethical video generation. Consequently, having set out as a deepfake detection solution provider in 2021, the AI startup also shifted its focus towards developing and monetising other GenAI applications.
Employing deep-learning models with substantial visual inputs, Kroop AI developed a text-to-video generation platform using digital avatars. Unlike platforms such as OpenAI’s Sora, Kroop AI’s generative platform allows users to animate avatars and create videos in various languages simply by inputting text.
“Our platform is language agnostic, meaning you can choose any language,” explained Joshi. “Once selected, you provide the script as input, which is then converted to the desired audio and used to generate a video where the avatar narrates the script.”
Kroop AI’s video generation platform supports over 25 languages, including English, French, Korean, Arabic, Hindi, Tamil, Telugu, and Malayalam.
Currently, the startup serves clients in the ecommerce and pharma industries. Besides, Kroop AI has found its biggest use case in the Banking, Financial Services and Insurance (BFSI) sector.
For instance, many insurance companies are using the Kroop AI platform to improve communication with their various stakeholders such as agents and customers.
Using its generative video technology, some companies have also launched new products and marketing campaigns. Flipkart’s furniture arm also adopted Kroop AI’s platform to convert its furniture catalogues to video formats from text format.
Meanwhile, the startup is yet to start monetising its deepfake detection tool. However, it is currently in talks with a few cybersecurity companies for the deployment of the technology.
Kroop AI’s Monetisation Strategy
Kroop AI’s generative platform provides services through three pricing tiers: $12 per month, $60 per month, and a customisable option.
The $12 plan offers 15 free minutes and access to a limited number of languages and avatars, while the $60 plan includes 80 free minutes and access to all languages and avatars.
According to Joshi, most large enterprises opt for the custom option, where pricing is tailored to their specific requirements. Additionally, the platform offers a free 7-day trial period.
To date, Kroop AI has secured $230K in funding from investors such as 100X.VC and LetsVenture, along with support from several angels and family offices. The company has not disclosed its monthly or yearly revenue.
However, it anticipates that in the coming years, its deepfake detection tools will surpass its generative platform as the primary revenue driver for the company.
The Road Ahead
Currently, Kroop AI is experiencing a surge in inquiries for its deepfake detection platform. According to Joshi, an increasing number of banks are approaching cybersecurity companies, which then turn to Kroop AI for assistance.
“With the advancement of tools, anyone can create highly realistic facial manipulations, potentially using another person’s image as their identity. Unfortunately, there’s often no deepfake detection layer before biometric verification, posing a significant risk. “This has led to a growing demand for deepfake detection tools from various banking sectors,” Joshi said.
Meanwhile, Kroop AI is eyeing expansion into regions such as the EU, MENA, and APAC. Particularly, banks in the APAC region have shown significant interest in its deepfake detection tool.
We must note that Kroop AI operates in a highly cost-intensive industry that requires high-end graphics processing unit (GPU) machines for data training. The startup is also looking to raise a fresh funding round by 2024 end.
