Bridging Tradition and Technology: Natya.AI for Bharatanatyam Dancers
Natya.AI is an app designed specifically for Bharatanatyam dancers, offering real-time voice and visual feedback during practice sessions. Developed by Yukti Arora as part of her thesis, Democratizing Bharatnatyam: Investigating the Interplay Between Dance and AI, this project explores how Artificial Intelligence can bridge the gap between tradition and technology, making the 2,500-year-old Indian classical dance form more accessible to today's dancers.
“80% of dance is practice and 80% of practice is correction.”
- Sri Thina, Bharatnatyam Performer & Teacher
Natya.AI uses computer vision to track the dancer's movement in real time, and data from benchmark Bharatanatyam performers are used to provide feedback on posture and timing. Using natural language processing (NLP), the app can interact and live-prompt the dancers during their practice sessions. It then provides a summary of feedback and splits the entire video into short snippets, making it easier for the dancers to find areas to work on and list the movements that need work. These movements then are added to a 'Movement Library' where a dancer can track their movement over time.
From her research survey with 300+Bharatntaym dancers, Yukti learned that 97% record themselves during their practice and consistently found the following three challenges:
Practice happens asynchronously: there is no live feedback from their Guru.
Review Fatigue: for every 5 minutes of practice, dancers spend up to 10 minutes scrubbing to find their mistakes.
There is also a lot of back and forth: setting up the camera, playing music and rewinding and redoing all of this when they stop in between.
Natya.AI was born out of the insight, that “80% of dance is practice and 80% of practice is correction” as quoted by Sri Thina, which made Yukti look further into the practice time of dancers. A survey further suggested that 97% of dancers record themselves while practicing to visualize their movements in space. However, recording & reviewing their practice is cumbersome, time consuming, and breaks the flow of practice. Dancers also experience mental fatigue while reviewing their practice videos. They miss the presence of their Guru who would live-prompt them if they made a mistake.
Natya.AI enhances the dancers’ practice time by providing real-time voice and visual feedback, mimicking the “in-class” experience. Different practice modes allow dancers to auto rewind and repeat parts of the practice that need improvement, decreasing interruptions and making the recording time feel like plug-and-play. Natya.AI also organizes the entire practice session into “splits” of good and bad areas, reducing reviewing time by 50%.
Before beginning the practice, the dancer chooses between three different modes that suit the different mental models the dancer has while practicing—such as ‘Focus mode’ where the AI is silently taking notes.
Through voice and gestures, the dancer can also have a conversation with the AI companion—helping with pauses, replays, rewinds and even advice on movement. To further make the feedback more ‘actionable’ (feedback that the dancer can work on), Natya.AI lists movements that need improvement in the summary view and also captures them in the ‘movement library’.
Process & Research
Yukti shares that her research methodology focused on interviews with subject matter experts (SMEs) in Bharatnatyam and AI, as well as reviewing scholarly articles and books. "My research helped me understand the current landscape of Bharatnatyam, as well as a sense of AI capabilities and how it can aid in the practice of Bharatanatyam," she said. "I actively engaged with a meetup group called NYC Adavus, an incredible community of 350 Bharatanatyam dancers that meet once a week to practice together. I conducted a survey to understand the challenges faced by dancers." Yukti received invaluable insights about their practice time, review fatigue, and how they miss the "in-class" experience with their Guru.
While developing the Natya.AI app, Yukti began by outlining the app's key features and identifying the minimum viable product (MVP) that would allow her to test my hypotheses and demonstrate how the app works. Once she had a clear idea of the MVP, she sketched out the different sections of the app. "This was an essential step in the development process, as it allowed me to think through the various scenarios and modes the app would need to accommodate," Yukti shares of the process. "Sketching also helped me think about the voice interfaces and interactions central to the app and the overall user experience."
As Yukti continued to develop the app, she found sketching was an invaluable tool for getting feedback on its usability. While high-fidelity wireframes helped get input on the concept, users felt more comfortable providing feedback via paper sketches. This allowed her to make changes and refine the app based on real user feedback, which was essential in ensuring that the app was as effective as possible.
To learn more about Yukti Arora's work, take a look at the other thesis projects that make up Democratizing Bharatnatyam: Investigating the Interplay Between Dance and AI. Find more of her design work at yuktiarora.com.