Youngster uses AI to recreate voices of musicians from K-Town

CHENNAI: Working as a manager at an Accountancy firm in the United Kingdom, Teejay-Sajanth Sritharan, a Srilankan Tamil, says he can’t be seen without headphones, which signifies his love for music. “In my company, usage of Artificial Intelligence (AI) is a regime.

That is when I started exploring the data models and wanted to try something in connection with music,” starts Teejay, an ardent musicophile.

Teejay produces voice models using AI, which models an individual’s voice that is trained to match a prominent singer’s voice for a specific song. For example, one of Teejay’s works includes recreating the Nee Kavithaigala song, originally sung by Pradeep Kumar, which was later matched to perfection with the late legendary singer SP Balasubrahmanyam’s voice.

“It is quite complicated. The process starts with extracting the voice from the sample. It is then cleansed to avoid noise and distortion. The sample is then uploaded to the system to separate every element of the song, starting from the vocal to instruments like piano, drums, guitar, and flute among others.

The segregated voice is trained through AI. The process is called e-poches, which means training over a period of time,” explains Teejay. He also adds that it could take up to 10 hours to train and develop a voice.

Concerned by the fact that AI influence was common in Hollywood and not in the Tamil film industry, Teejay wanted to create and blend his creativity with technology. “After delving deep into the process, the respect I have for music composers has increased because of the level of work involved in composing just one song. It is insane,” he shares.

Not only this, Teejay says that this concept can be used for dubbing in different languages. “If a Tamil film is released in Hindi, mostly the voice of the artistes will be dubbed by someone else. But using this technique, the voice of the actors can be trained to speak any language,” he suggests.

Teejay has created voice models for actors like Vijay, Dhanush, Silambarasan TR, Ajithkumar, and musicians SP Balasubramanyam, AR Rahman, Anirudh Ravichander, Shreya Ghoshal, Yuvan Shankar Raja and Pradeep Kumar, among others.

Talking about the criteria with which he chooses the song and recreated voice, Teejay shares that he takes suggestions and requests from social media users. He makes voice models in Telugu, Malayalam, and Hindi as well. “Apart from the many advantages, if this AI is used in the music industry, singers can create their own voice models and get royalties for the same.

For me, the Tamil industry should be the pioneer in introducing this technology,” remarks Teejay, who is an ardent fan of SP Balasubrahmanyam and Anirudh. Music composer Sean Roldan has reshared a few works of Teejay on social media and music label Think Music has appreciated his creations.

Teejay is expected to work on a couple of projects for the Tamil film industry in the coming months. “The whole purpose of creating these voice models is to recreate the lost iconic voices from the industry,” Teejay signs off.