Listen to this story
A couple of weeks ago, Netflix released a trailer for an anime project named “Dog and Boy”. The entertainment platform claimed that since the anime industry has a labour shortage, they used image generation technology to generate background images of all three-minute video cuts.
While the platform faced severe criticism from anime fans for using AI to generate images rather than hand-drawing them—as is conventional—the entertainment industry is all up for embracing the new tech. For instance, Eros Now is already using AI to stream an entire range of Eros Now films and originals with automated subtitles.
Similarly, Amazon Prime uses ML to enhance the video quality, as it might be harmed by flaws introduced during recording, encoding, packing, or transmission. Not only Amazon but most of the video streaming services use ML in a similar manner to enhance their video quality.
Subscribe to our Newsletter
Join our editors every weekday evening as they steer you through the most significant news of the day, introduce you to fresh perspectives, and provide unexpected moments of joy
To know more about how Indian streaming services are using technology to enhance the productivity of their employees, Analytics India Magazine got in touch with Kishore AK, CTO of Zee5 at the Big CIO Show in Bengaluru.
Zee5, according to AK, is using AI in its content creation process along with customer-facing activities. While AI is already used for subtitles and closed captioning, Zee5 is now experimenting with using generative AI to help its content creators augment their work. “I’m sure my development team is using Github Copilot at some scale,” AK explained. However, he also believes that while using Copilot at the initial stage and for training the freshers can be encouraged, once the IP algorithms are formed, the use of such platforms might pose challenges of data leaking.
Additionally, AK goes on to say that generative AI is also being used to help content creators generate scripts for upcoming shows. Although he declined to reveal the name of the shows. “If everything goes according to plan, we might continue utilising LLM Models in script writing”.
However, not all are happy with AI being introduced in the video streaming industry—particularly the employees whose jobs are most likely to get affected. For instance, in a media interaction, Eros Now CEO Ali Hussein revealed that prior to the introduction of AI in the subtitling industry, it used to take up to 48 hours for each movie to be subtitled correctly—something that can now be done now in matter of seconds.
Similar was the case with Netflix’s “Dog and Boy” video clip. While anime fans were adamant about using artists to create the background, the artists themselves are costly to the industry. As per industry experts, experienced artists earn around $1500~3000 a month in the anime industry—a cost that Netflix is aiming to reduce at their end.
Recommendation system and Privacy
When asked if Netflix’s recommendation engine is the best in the industry, AK was of the opinion that it can not be compared. “Our audience is diverse and varies depending on the language and region”.
He further explained that to improve recommendations, the platform initially relied on English metadata but it is now starting to extract features from video content. “You see, to build a recommendation system for the English language is far easier than the local Indic languages as it depends a lot on the data upon which a particular ML model is trained.”
For most of its local content, AK says Zee5 is still translating it first to English and then feeding the data to the algorithm. “Once the natural language models are capable enough to understand local Indic languages, we might see an improvement”.
Furthermore, he said that platforms like Spotify have almost all songs present on the platform, or like Netflix which has huge data at its disposal.
“Zee5 has a niche segment of that content, which further makes it difficult to get a recommendation as powerful as other market players.”
While discussing the recommendation system, the CTO also highlighted the importance of protecting user data and privacy. He explained that Zee5 has standardised ways of executing tasks when it comes to user data management, such as looking at the user’s whole data lifecycle and what happens when it’s moving around.
After a certain period of time, AK explains, Zee5 starts deleting user data holistically and has implemented deep encryption levels and access based on roles to ensure comprehensive security. Additionally, the platform allows users to control their own data, giving them the option to delete it or make it completely anonymous.