Exclusive Interview with Subhabrata Debnath, Co-Founder and CTO of NeuralGarage

NeuralGarage is a pioneering tech company at the forefront of neural network innovation. The initiation of generative AI indicates an expansion of its product capabilities within the AI and Computer Vision industry. The company's focus on enhancing the viewing experience for non-native audiences suggests that it is innovating and potentially increasing its market reach. Furthermore, the fact that NeuralGarage is being discussed in the context of technology news indicates that it is gaining visibility in its sector. Analytics Insight has engaged in an exclusive interview with Subhabrata Debnath, Co-Founder and CTO of NeuralGarage.

Kindly brief us about the company, its specialization, and the services that your company offers.

NeuralGarage has built VisualDub to visually change the lip and jaw movements of actors in dubbed content, making them look real and authentic.

With what mission and objectives was the company set up? In short, tell us about your journey since the inception of the company.        

NeuralGarage was formed to serve the media and entertainment industry and make content consumption seamless by using the power of Generative AI technology.

Me(Subhabrata), Anjan, and Subhashish have known each other since 2013. We were batchmates at IIT Kanpur. After passing out(2015), we co-founded a facial recognition company called VisageMap, which was acquired by a large US-based company called FaceFirst. After that, we worked together for nearly 5 years at FaceFirst, spearheading their entire research and gaining considerable experience in Generative AI algorithms.

NeuralGarage is the second company we are building together along with Mandar who has over 20+ years of experience in the media industry.

Anjan loves to consume Korean content, and during COVID-19, his consumption went even higher. However, he realized that while dubbed content makes content accessible, it also causes visual dissonance, which impedes the content consumption experience as the dubbed audio we hear is not in sync with the mouth movements of the actors. The three of us got together and decided to solve this.

We had reached out to Mandar through a common connection to understand the value it brings to the media industry and whether it is something the industry will be willing to pay for.

After visiting 30+ studios and meeting with various producers and directors, we realized the industry desperately needs the solution, and hence, we formed NeuralGarage in 2021.

Mention some of the awards, achievements, recognitions, and feedback from clients that are notable and valuable for the company.   

The following are the recent achievements and accolades that we have received:

●      Being featured on Times Techies, TOI edition of 24th May 2024

●      One of the top 26 startups in India selected as a part of the Nasscom Generative Foundry

●      NeuralGarage placed 2nd in PitchFest at TechSparks 2023 - YourStory

●      NeuralGarage part of Tech30 2023 - YourStory

●      One of the top 60 GenAI startups in India by Inc42

●      30 startups to watch for by Inc42

●      Featured on YourStory

●      One of the top startups being selected by Amazon AWS is Elevate across India and Korea

●      Nasscom Emerge50 awards(Top 50 emerging startups in India)

What is your biggest USP that differentiates the company from competitors?  

VisualDub creates studio-quality LipSync, which retains the original quality of the video and produces natural, non-AI-ish outputs.

Please tell us about the products/services/solutions you provide to your customers and how they benefit from them.

At present, we provide 3 solutions to advertisers, media companies, and creative agencies:

Visual Multilingual LipSync: When an advertisement video is shot in one language(say Hindi) and dubbed into other languages(say Tamil, Telugu, Malayalam, Bengali, etc.), we can recreate the video in these languages so that it looks as if the actors in it are actually speaking that language visually.

Visual Dialogue replacement: We can create multiple videos by shooting just one video with an actor and using different audio files. With the help of VisualDub, we can change the lip and mouth movements to make it seem as if the actor is now visually speaking the different audio.

Visual Hyper-personalization: From one video message where an actor is speaking, we can create millions of videos at scale, each personalized to a specific user by changing a few variables in the message(say the name, age, etc.), such that it seems that the video message was specifically meant for them.

We have onboarded over 16+ large brands that have been using our solution for the past 86 months. Some of the names include Amazon (Fresh, Daily, Fashion, Pay), Britannia, Coca-Cola, Dream11, HP, UltraTech Cements, Eno, and Microsoft US. The uses have primarily been Dialogue replacement and Multilingual lip sync.

Could you highlight your company's recent innovations in the AI/ML/Analytics space?     

We have built our own proprietary models using Generative AI specifically for video. We have already filed for a patent. Our model can change a person's lip and jaw movements using audio as an activation at studio quality while retaining the quality of the original video.

What is the skill set that you specifically look for while hiring them at an initial stage?

We highly value the following skills: approach to a problem, self-criticism, ownership and responsibility, and domain knowledge (in probability, vector algebra, and statistics).