The "Desifakes" movement relies on sophisticated machine learning models, primarily RVC (Retrieval-based Voice Conversion). These models are trained on hours of isolated vocal tracks from specific artists—such as Arijit Singh, Shreya Ghoshal, or the late Sidhu Moose Wala. Once the model captures the unique timbre, vibrato, and pitch range of the singer, it can be applied to any melody. Fans use this technology to create:
The popularity of Desifakes songs is driven by nostalgia and curiosity. Social media platforms like TikTok, Instagram, and YouTube are flooded with "What If" scenarios. For many fans, hearing a beloved deceased artist "sing" a new song provides a sense of emotional closure or continued connection. In the fast-paced world of Desi pop culture, these AI tracks often go viral faster than official releases because they tap into established fanbases with highly specific, imaginative content. Ethical and Legal Minefields
Cross-genre covers: A Punjabi folk singer performing a Hollywood pop hit.
While the technology offers a playground for creative mashups, it also raises significant ethical and legal questions regarding artist consent, copyright, and the future of human creativity in the digital age. The Mechanics of Desifake Audio