Kuku FM, a popular audio content platform backed by Google and Nandan Nilekani’s Fundamentum Partnership, is harnessing the power of generative AI to revolutionise how stories are created, produced, and consumed. This transformation is spearheaded by Kunj Sanghvi, the VP of content at Kuku FM, who told AIM that generative AI is part of their everyday work and content creation.
“On the generative AI side, we are working pretty much on every layer of the process involved,” Sanghvi explained. “Right from adapting stories in the Indian context, to writing the script and dialogues, we are trying out AI to do all of these. Now, in different languages, we are at different levels of success, but in English, our entire process has moved to AI.”
Kuku FM is leveraging AI not just for content creation but for voice production as well. The company uses Eleven Labs, ChatGPT APIs, and other available offerings to produce voices directly.
“Dramatic voice is a particularly specific and difficult challenge, and long-form voice is also a difficult challenge. These are two things that most platforms working in this space haven’t been able to solve,” Sanghvi noted.
In terms of long-form content moving to generative AI, Kuku FM also does thumbnail generation, visual assets generation, and description generation and Sanghvi said that the team has custom GPTs for every process.
Compensating Artists
AI is playing a crucial role in ensuring high-quality outputs across various languages and formats. “In languages like Hindi and Tamil, the quality is decent, but for others like Telugu, Kannada, Malayalam, Bangla, and Marathi, the output quality is still poor,” said Sanghvi.
However, the quality improves every week. “We put out a few episodes even in languages where we’re not happy with the quality to keep experimenting and improving,” Sanghvi added.
Beyond content creation, AI is helping Kuku FM in comprehensively generating and analysing metadata. “We have used AI to generate over 500 types of metadata on each of our content. AI itself identifies these attributes, and at an aggregate level, we can understand what makes certain content perform better than others,” he mentioned.
One of the most transformative aspects of Kuku FM’s use of AI is its impact on creators. The platform is in the process of empowering 5,000 creators to become full-stack creative producers.
“As the generative AI tools become better, every individual is going to become a full-stack creator. They can make choices on the visuals, sounds, language, and copy, using AI as a co-pilot,” Sanghvi said. “We are training people to become creative producers who can own their content from start to end.”
When asked about the competitive landscape such as Amazon’s Audible or PocketFM, and future plans, Sanghvi emphasised that AI should not be viewed as a moat but as a platform. “Every company of our size, not just our immediate competition, will use AI as a great enabler. AI is going to be commoditised; everybody will have access to the tools. What will remain crucial is the talent pool you have – the storytellers,” he explained.
Everyone’s a Storyteller with AI
In a unique experiment blending generative AI tools, former OpenAI co-founder Andrej Karpathy used the Wall Street Journal’s front page to produce a music video on August 1, 2024.
Karpathy copied the entire front page of the newspaper into Claude, which generated multiple scenes and provided visual descriptions for each. These descriptions were then fed into Ideogram AI, an image-generation tool, to create corresponding visuals. Next, the generated images were uploaded into RunwayML’s Gen 3 Alpha to make a 10-second video segment.
Sanghvi also touched upon the possibility of edge applications of AI, like generating audiobooks in one’s voice. “These are nice bells and whistles but are not scalable applications of AI. However, they can dial up engagement as fresh experiments,” he said.
Kuku FM is also venturing into new formats like video and comics, generated entirely through AI. He said that the team is not going for shoots or designing characters in studios. “Our in-house team works with AI to create unique content for video, tunes, and comics,” he revealed.
Sanghvi believes that Kuku FM is turning blockbuster storytelling into a science, making it more accessible and understandable. “The insights and structure of a story can now look like the structure of a product flow, thanks to AI,” Sanghvi remarked.
“This democratises storytelling, making every individual a potential storyteller.” As Sanghvi aptly puts it, “The only job that will remain is that of a creative producer, finding fresh ways to engage audiences, as AI will always be biassed towards the past.”