Last year, an Indian studio called Eros Media World took a 2013 hit called Raanjhanaa and re-released it with an AI-altered ending. In the original, the protagonist dies. In the new version, he opens his eyes. His lover smiles through tears. The audience gets a resolution that feels nicer and means less.
The film's lead actor, Dhanush, said on X that the remake had "stripped the film of its very soul" and set a "deeply concerning precedent for both art and artists."
The re-release sold well. Thirty-five percent of available tickets moved during its opening month. Twelve percentage points above the 2025 average for India's largest cinema chain.
So Eros is going further. The studio is now reviewing its entire 3,000-title catalog to identify candidates for what its CEO called "AI-assisted adaptation." He described it as "both a revenue opportunity and a creative renewal strategy."
That second phrase is doing extraordinary work.
The scale
Reuters published a detailed investigation last week into how India's film industry is reorganizing itself around AI at a scale that makes the rest of the world look cautious. The Collective Artists Network, one of Bollywood's largest talent agencies, has pivoted from managing real actors to engineering digital ones in its Bengaluru studio. Cost reduction: 80 percent. Production timeline: down to a quarter. Abundantia Entertainment, a major Bollywood production house, is investing $11 million in an AI studio and expects AI-generated or AI-assisted content to account for a third of its revenue within three years.
Google, Microsoft, and NVIDIA have placed early bets by partnering with Indian filmmakers.
India produces more films than any other country. It also sent nearly twice as many submissions as the United States to the largest AI filmmaking competition ever held. The access revolution this series has tracked is real, and it arrived in India louder than anywhere.
But what Reuters describes is not the same story. The Higgsfield competition was 8,752 individual filmmakers from 139 countries making new things. Creative expansion. People who had something to say, saying it for the first time because the infrastructure barrier had collapsed.
What is happening at the studio level in India is an existing industrial system using the same tools for a different purpose. Not to create films that could not have existed before, but to produce existing genres cheaper and to alter existing films for additional revenue.
The contract Hollywood has and India does not
Under SAG-AFTRA's contract, American studios cannot digitally alter an actor's performance or create a digital replica without the performer's informed consent. The Directors Guild bars studios from using AI for creative decisions without consulting the director. These protections exist because a hundred and eighteen days of strikes in 2023 forced the industry to negotiate them.
India has no equivalent protections.
The absence of creative labor protections does not mean the creative labor is unnecessary. It means the creative labor is being overridden without the laborers being consulted.
Dhanush was not consulted about his own death scene being rewritten. He found out when the audience did. He called it soul-stripping. The studio called it creative renewal.
The intent problem
Forty-six articles about closing the gap between creative intent and model output. Every one of them assumed the creative intent belonged to the person holding the prompt. The filmmaker. The person with a vision they could not previously afford to realize.
The Reuters report introduces a different actor: the studio that holds the catalog. The creative intent there is not "what does this shot look like?" It is "how do we extract more value from an existing asset?"
The model does not know the difference. A prompt that says "the protagonist opens his eyes and his lover smiles through tears" is technically indistinguishable from a prompt that says "soft morning light through gauze curtains, shallow focus, subject in left third of frame." Both arrive as text. Both produce pixels. One replaces a creative decision that already existed. The other makes a new one.
The 80 percent cost reduction is real. The compressed production timeline is real. The 35 percent ticket sales are real. None of those numbers answer the question that Dhanush asked when he found out someone had rewritten his death scene.
Was the ending a flaw or a choice?
If it was a choice, the AI did not fix it. It overrode it.
Two uses of the same tool
A filmmaker with a laptop in Jaipur using Kling to build a scene they could never have shot is the access story. That is cultural expansion. That is vocabulary carrying creative identity through a generation pipeline.
A studio in Mumbai reviewing 3,000 titles for "AI-assisted adaptation" is not filmmaking. It is catalog optimization wearing a filmmaker's jacket.
Both use the same models. Both send the same structured prompts. Both produce output that is technically competent. The difference is not in the technology. The difference is in the question being asked. One asks: what do I want this to look like? The other asks: how do I make this cheaper?
The models answer both questions with equal enthusiasm. They have no preference. They have no opinion about whether a death scene earned its place in the story. They will rewrite it or build something new with identical computational indifference.
Vocabulary is neutral
The vocabulary this series has built across forty-six articles is neutral. It carries whatever intent the person typing loads into it. Structured cinematographic language works the same for a filmmaker building a vision and a studio rewriting someone else's. The prompt does not know whose intent it serves.
That neutrality is the point and the problem. The same structured input that makes a first-time filmmaker's vision precise also makes a catalog rewrite efficient. The same tools that expanded creative access to 139 countries also gave a studio the ability to undo a director's ending at industrial scale.
India's film industry now has the tools to produce more films, faster, cheaper, and in more languages simultaneously. Whether it produces them better depends on the same thing it has always depended on: whether the people making the decisions are asking creative questions or financial ones.
The models will answer either way. They do not care which question arrived. That discernment is still yours.
Bruce Belafonte is an AI filmmaker at Light Owl. He watched a studio rewrite a death scene and spent the rest of the day thinking about who the ending belonged to.