Aniket Bera, director of the Ideas Lab at Purdue University, has worked on two very different projects in the field of AI: the restoration of an 1899 film fragment, believed to be India’s oldest surviving footage, and an earlier AI-based experiment with Satyajit Ray’s Pather Panchali.
“The AI softens the shadows and contrasts that were central to the mood of the film. The AI doesn’t understand symbolism, it only guesses at patterns,” he says. Bera says each step required human review to ensure the result was true to the original. “AI often hallucinates details, ‘improving’ things by changing the visual language. With this, we risk rewriting history.”
For Mukherji, AI allowed him to realize his cinematic vision. How else could he have chosen two deceased actors? AI recreated Uttam Kumar’s voice throughout Oti Uttam. However, he emphasizes that the project still depended heavily on human input: for writing the scenario, collecting archive images, seeking legal authorizations and controlling the results of the AI.
AI tools are evolving rapidly, creating a host of regulatory and ethical questions. Mukherji calls for optimism. “Instead of panicking, humans should familiarize themselves with AI,” he says. “Tame it, master it, and harness it. It’s not an android-like monster trying to swallow up your creativity. It aids creativity, doesn’t replace it.”
And yet, for others, the limits of AI remain clear. Chandu now shares his on-set learnings in the classroom: he teaches a university course on AI in cinema. In one module, he encourages students to make two films: one using ChatGPT and AI video tools, and the other entirely with traditional techniques.