The short answer. In India in 2026, studio rate cards start a 30-second AI ad film at ₹21,000 to ₹75,000, while AI films made to a brand’s television standard are reported at roughly ₹4 to 12 lakh, against about ₹25 lakh for a broadcast live-action TVC in our own guide. Choose AI when nothing on screen has to be real, live action when the promise is a real person or product, and hybrid in between.
Most of what you will read on this question is written by studios that sell only one of the three. We make all three. This year we made Air India’s 102-second AI film with McCann, 32 fully AI founder films for Razorpay’s IPL season, and hybrid films for Malaysia Airlines and Meta, alongside the live action we have shot since 2010. Our vivo X300 Pro launch film, directed by Atul Kattukaran, won a Bronze, shared with Old Monk Studio, in the AI-Integrated Live Action Film category at the Good Ads Matter Awards 2026. So this is the producer’s version: what each route costs, where the money goes, and how to decide for your brief, first at the brief and then one shot at a time.
Updated 10 October 2026 with a test for deciding at the brief, our own AI and hybrid films from this year, and what India’s AI ads and their audiences show. The cost figures were rechecked on 9 October.
Key takeaways
- Two different products are sold as "AI ad films" in India: studio-made digital films whose 30-second entry prices run from ₹21,000 to ₹75,000, and brand-grade films reported in the trade press at several lakh rupees.
- Published savings claims run from 30% to 95%. They disagree because each one is measured against something different: a shoot, a CG film, a fantasy epic.
- AI does not delete the budget. It moves money out of shoot days and into look development, generation, post and compliance. The footage itself is the cheap part: about ₹11,500 to ₹34,600 for a 30-second film at Google’s published Veo price.
- Decide at the brief first: what the film must prove, who must be on screen, how many films it is, and how early you can sign off each scene.
- Then decide shot by shot. Shoot the people, the product and the proof. Generate the world, the scale and the versions.
- AI is fast only after the stills are locked. In our pipeline a short film then takes about a day of generation, and a character changed afterwards resets the scene.
- Since 20 February 2026, realistic AI footage carries a label on Indian platforms, and ASCI’s final AI guidelines, released on 29 September 2026, ban fake testimonials and exaggerated results outright.
What does an AI ad film cost in India in 2026?
An AI ad film in India costs anywhere from under a lakh to the low tens of lakhs, because two different markets use the same three words.
The first market is the published price list. On 5 October 2026 we read the rupee prices published by 20 Indian AI video studios and vendors, 100 price points in all. Six of them publish a rate card for a 30-second film, and their entry prices run from ₹21,000 to ₹75,000, with a median of about ₹45,000. Their top ad-film packages reach ₹1.25 to 1.5 lakh, and some add GST on top. We rechecked the six rate cards on 9 October; none had changed. These are prices for digital and social films with a fixed number of revisions, made quickly, often by small teams.
The second market is what brands report paying for AI films made to the standard of their television work. Nobody publishes those as a price list. They surface in trade interviews, and they sit higher. Carol Goyal of Aesthetic Intelligence Lab told Storyboard18 in July 2025 that AI-led campaigns which would cost ₹50 to 60 lakh conventionally can come in at ₹8 to 12 lakh, depending on scale and scope. Sanjeev Jasani, formerly COO of Cheil India, told exchange4media in August 2025 that a TVC which cost ₹20 to 25 lakh can now be made at similar quality for about ₹4 to 5 lakh. AI director Akshay Srivastava’s rule of thumb, reported by BestMediaInfo in June 2026, is to budget the script as if it were a real shoot and then cut that to a third.
Films at that standard are already on Indian television. Tata Gluco+'s 2025 film, conceived by VML and made entirely with AI by Rediffusion’s RAIDS studio, went live in media including the IPL, and TVS billed its Jupiter ad as the first fully AI-generated TV ad for a scooter brand in India when it debuted the same season.
For the live-action side, here are the tiers from our guide to ad film production cost in India, next to the AI numbers above.
| Route | What it costs in India, 2026 | What you usually get | Source |
|---|---|---|---|
| AI, studio rate cards | ₹21,000 to ₹75,000 to start for 30 seconds (median about ₹45,000); top ad-film packages ₹1.25 to 1.5 lakh | Digital and social films, fixed revisions, small team | Six published rate cards, read 5 Oct 2026, rechecked 9 Oct |
| AI, brand-grade | About ₹4 to 5 lakh for what would be a ₹20 to 25 lakh TVC; ₹8 to 12 lakh for what would be a ₹50 to 60 lakh campaign | Consistent characters and product, full edit, grade and sound, TV delivery | exchange4media, Aug 2025; Storyboard18, Jul 2025 |
| Hybrid | Set by the shoot days you keep, plus compositing in post | Real people and product on set, generated worlds and extensions | Our view; Studio Blo sells a single AI shot inside a live film |
| Live action, digital film | ₹8 to 25 lakh | A proper director, DOP and crew, one or two shoot days | Our cost guide |
| Live action, national TVC | ₹25 to 60 lakh | Broadcast standard, two to four shoot days, a crew of forty to eighty | Our cost guide |
| Live action, celebrity | Production from about ₹75 lakh, before the fee | Everything above, built around one person’s calendar | Our cost guide |
All production cost only. Media is a separate, usually larger, budget. GST at 18% is extra on most quotes.

Why do "AI is cheaper" claims range from 30% to 95%?
Every savings figure in circulation is true for the film it was measured on, and each was measured against something different.
| Claim | Who said it | Compared with |
|---|---|---|
| An AI film may cost 5 or 10% of a shoot, but will look only 5 or 10% as good | Devaiah Bopanna, Moonshot, in BestMediaInfo, Jun 2026 | A live shoot, with a warning attached |
| ₹8 to 12 lakh for an AI-led campaign that would cost ₹50 to 60 lakh conventionally | Carol Goyal, Aesthetic Intelligence Lab, in Storyboard18, Jul 2025 | A conventional campaign |
| 15 to 20% of a similar CG-heavy film | Kaushik Vedula, Emami, on DermiCool Warriors, in Storyboard18, Jul 2025 | A CG film, not a shoot |
| One-fifth the cost and a quarter of the time | Rahul Regulapati, Galleri5, to Reuters, Apr 2026 | Mythology and fantasy filmmaking |
| About ₹4 to 5 lakh against ₹20 to 25 lakh | Sanjeev Jasani, ex-Cheil India, in exchange4media, Aug 2025 | A TVC "at similar quality" |
| A third of the shoot budget | Akshay Srivastava, AI director, in BestMediaInfo, Jun 2026 | The same script budgeted as a shoot |
| A 30 to 40% saving, and not plug-and-play | Aniket Sakpal, Agency09, in Storyboard18, Jul 2025 | Live production |
| Not always cheaper, especially with celebrities or real locations | Harikrishnan Pillai, TheSmallBigIdea, in Storyboard18, Jul 2025 | Any production with talent |
Put the seven claims that carry a number on one scale, and four of them land in the same place: an AI film at roughly 13 to 25% of what it replaced. The two ends sit far apart: 5 to 10% of a shoot at one end, 60 to 70% of live production at the other.
The extremes tell you more than the middle. The biggest savings compare AI against the most expensive way of making the same pictures: a CG world, a battle, a god. The smallest come from people who still pay for the edit, the grade, the sound and the voice. BestMediaInfo’s own reporter concluded that real savings are less dramatic than the 80 to 90% figures usually quoted.
Vendor pages add a second distortion. Where they say what they compare against, most price a basic TVC at ₹2 lakh to ₹5 lakh, a fifth or less of where our national TVC tier starts. DigiStreet, one of those vendors, warns on its own cost page that 90% savings claims usually assume the creative labour costs nothing.
The useful reading is simpler. AI saves the most where the camera was the expensive part, and the least where people were. If your budget is mostly shoot days, sets, travel and locations, AI can cut it hard. If it is mostly talent, edit and finish, it will not.
The production figures brands and studios have published bear this out, from one person working for two to three days to 37 people over 56 days. We collected them in the state of AI commercials 2026.
The 2027 evidence points the same way: speed to market has overtaken cost savings as UK advertisers’ main reason for using AI in production, and seven in ten agencies told D&AD they have not changed what they charge. We graded that and eleven other claims in our 2027 advertising trends guide.

Where does the money go in an AI ad film?
AI does not delete an ad film budget. It moves money between the four buckets every film already has, and it adds a few lines a live shoot never needed.
Our cost guide sorts every rupee of a live-action film into pre-production, production, post-production and talent. Here is what each route does to them.
| Bucket | Live action | Hybrid | Fully AI |
|---|---|---|---|
| Pre-production | 10 to 15% of the budget: treatment, casting, recce, permissions | The same, plus planning which plates to shoot for the generated shots | Grows: reference boards for every character, costume, location and product, approved before anything is generated |
| Production | The largest bucket: shoot days, crew, camera, lights, sets, locations | Shoot days stay for the people and the product; some sets and locations move to generation | Shoot days disappear; generation replaces them: credits, GPU time and AI artists working in layers |
| Post-production | Edit, grade, sound, music, VFX, versions | Grows: compositing generated shots into real footage and matching the light | Stays largely traditional: edit, clean-up, grade, sound, music, upscaling for delivery |
| Talent | Actors, usage terms, celebrity entourage | Unchanged for everyone on camera | Shrinks for invented people; returns as voice actors, licensed music and consent fees for any real likeness |
| New lines | None | Labelled and clean masters; consent records | Labelled and clean masters; consent records; a log of the human decisions behind the generations |
The trade interviews back the table. Studio Blo’s Dipankar Mukherjee told Social Samosa in June 2026 that post-production stays largely traditional in time and cost, and that VFX bills fall but do not disappear. Carol Goyal told Storyboard18 that the savings often move into prompt engineering, voice licensing, GPU compute and human quality checks. Srivastava puts generation credits alone at ₹40,000 to ₹1,00,000 per finished minute, with an editor at ₹25,000 to ₹30,000 a project and a colourist and sound team on top.

What does the footage itself cost?
Google prices its Veo 3.1 video model at $0.40 a second for 1080p footage with sound, and its cloud catalogue lists that at about ₹38 a second, as of October 2026. AI films are generated the way live films are shot, with many takes for every one used: AJIO’s 2025 AI campaign made more than 200 clips for a 22-scene film, and on our Air India film more than 600 generations were reviewed for the generated shots in a hundred-shot cut. At ten seconds generated for every second kept, a 30-second film needs 300 seconds of footage, about ₹11,500 at Google’s price. At thirty to one it needs 900 seconds, about ₹34,600. Kling 3.0’s API comes to roughly ₹4,800 and ₹14,500 for the same two ratios.
In a brand-grade film of ₹4 to 12 lakh, that footage is the cheapest line on the budget: even at thirty to one it is about 9% of a ₹4 lakh film and 3% of a ₹12 lakh one. The money goes to the people: the director, the AI artists, the editor, the colourist, the sound designer, and the days it takes to keep one face the same face for thirty seconds.
In a rate-card film at about ₹45,000, the same footage would take about 77% of the price. That suggests what the cheapest AI films economise on: fewer attempts per shot and cheaper model tiers, which is to say less choosing.

Why thirty AI films cost less each than one
An AI film has set-up costs that do not shrink with runtime: the approved look, the character and product references, and the days a team spends learning what this brand’s world should look like. On one 30-second film that set-up is heavy. Across a campaign it is spread thin.
The Indian examples that report the biggest gains are all volume stories. Storyboard18 reported that Cadbury Silk made more than 500,000 unique videos, and Mondelez India’s Nitin Saini said traditional production cannot match that scale. Carol Goyal said the Tata Gluco+ campaign delivered regional variants within a week. Our own clearest case is Razorpay’s IPL season: five days of set-up and a paid test film, then about one film a day, 32 films in all.
So ask a different question before you ask about price: how many films is this? If the answer is one, the savings are real but modest. If it is twenty versions in six languages, they are large.
AI, live action or hybrid: how do you decide for your brief?
Decide twice. First at the brief, before anyone writes a shot list: four questions tell you which routes are open. Then at the script, where five questions settle each shot. If one scene still will not settle, test it before you commit the budget.
Which four questions settle the route at the brief?
Most of the decision is made before the script exists, by four answers in the brief.
- What does the film have to prove? If the promise is that something is true, such as a product working, a result, a real person’s story or food as it really is, the proof has to be shot. ASCI’s final guidelines ban fabricated testimonials and exaggerated results even when they carry a label. If the promise is imagined, a world that does not exist or a feeling at the scale of a country, AI can carry it, often further than a camera can. Vivek Anchalia of Amazing India Stories, who makes AI films, told Goafest 2025 that AI suits montage ads with no emotional connection between two people.
- Who has to be on screen? A famous face, a founder or a real customer is shot, or generated only under written consent that covers scope, duration, fee and where the film runs. Producers abroad ask the same question. Andrew Loevenguth of Cartwright put it to LBB in April 2026 as: "Could we shoot the celebs separately and then use AI to put them in the spot?" Invented characters can be generated, but each one has to be fixed as a still and held through every angle.
- How many films is it? The set-up of an AI film costs the same at any volume: the look, the characters, the references, the pipeline. Over one film it is heavy. Over thirty it nearly disappears. Count formats and languages as films, because a vertical 9:16 version is usually a separate generation rather than a crop.
- How early can you sign off what each scene is? AI is fast only after the characters, costumes and locations are approved as stills. In our pipeline a 15 to 30 second film takes about a day of generation once its stills are locked, and a character changed after that point resets the scene. If your approvals take three weeks, AI will not make the film faster.
| What the brief says | Leans to | Why |
|---|---|---|
| The film must prove something real | Live action | Labels, and ASCI’s ban on fabricated proof |
| A famous face carries it | Hybrid: shoot the face, extend the world | Consent, fee, and a face viewers know too well |
| Invented people in a world that does not exist | AI | Nothing on screen has to be real |
| Many films, formats or languages | AI or hybrid | The set-up spreads across the volume |
| Sign-off on characters and locations will be slow | Decide on the other rows, not on speed | AI is fast only after the stills are locked |
The brief test. Most briefs say yes to more than one row. That is why most films end up hybrid.

Then which shots do you shoot, and which do you generate?
Once the script exists, run every shot through five questions, in this order.
- Does a real face or performance carry the shot? Shoot it. Even AI directors say so. Simon Dolsten wrote in LBB in July 2026: "You shoot the performance live, because nothing synthetic yet matches the micro-expressions that make a face read as human." Keeping one face consistent across angles is also what AI studios such as Studio Blo price as complexity, and a viewer who knows the face notices any drift first.
- Does the product have to be exactly right? Shoot it, or composite a real packshot. In Coca-Cola’s 2025 AI Christmas film the truck is the hero, and Dino Burbidge counted in The Drum that only two of its 11 truck scenes show the same truck. A generated product is always one prompt away from looking better than the real one, and ASCI prohibits exaggerated results outright.
- Is the shot impossible, unsafe or ruinously expensive to shoot? Generate it, or extend a real plate. A world, a crowd, a period street, a storm, a mythological army.
- Will the shot exist in many versions? Build it so AI can make the versions, and pay a person who knows each market to check every one. When Sprite made AI city versions of its 2026 summer film, one widely discussed observation, reported by Storyboard18, was that the Kolkata frame reportedly showed Sydney’s Harbour Bridge instead of the Howrah Bridge.
- Does the film’s promise depend on "this really happened"? Shoot it. Realistic AI footage of people or events now carries a label on Indian platforms, a label that says AI sits badly on a film whose whole claim is that something real happened, and in testing, disclosure has cut click-through rates by almost a third.
When every shot lands on "generate", make the film fully AI. When every shot lands on "shoot", make it live action and use AI where Ganesh Pareek put it in his BRIEF26 note for MediaBrief in March: for compression, meaning logistics, iterations and prep, and never as a substitute for taste or judgment. Most scripts land in between. That is hybrid.

What if one scene will not settle?
Test that scene before you commit the budget. A crowd that has to read as Indian, a label that has to stay legible, a face that must not drift: each is cheaper to test as one scene than to discover as a reshoot or a reset. The most committed AI producers say the same. PJ Accetturo, whose Genre.ai made Kalshi’s AI ad for the 2025 NBA Finals, told Wrapbook’s On Production podcast in September 2026 that "not all scripts should be AI". A one-location restaurant spot his team made could not hold the location or the actors’ performances, and he wished they had shot it at a diner for about $40,000.
A test can double as research. Kantar data cited by the Financial Express in September 2026 showed that 15% of the unfinished ads it tested in India in 2025 were GenAI animatics, and that GenAI ads performed at par with traditionally produced ones on several measures, including emotional response.
We sell this as a paid look-development and feasibility test of the one scene everyone says cannot be done, before full production, on our AI films page.
What did each route look like in our own work?
AI, one film at scale. Air India’s There is an Air About India, made with McCann and released in September 2026, follows India through a young cricketer’s eyes, from his father’s shoulders to a billion people cheering. It runs 102 seconds: about a hundred shots, the generated ones chosen from more than 600 generations and the rest licensed stock footage. Every generated shot passed a realism gate of twelve checks, from hands and eyes to weight, background people and morphing, and one wrong behaviour sent a shot back however good it looked. The hardest check is cultural. Ram Madhvani said in December 2025 that he could not yet use AI "as Indian as I’d like to", and every new shot on Air India was tested against six words: rooted, Indian, observed, layered, imperfect, human. The edit began while generation was still running, so each new prompt was written for the cut rather than for a pretty clip. The master carries its label: "Some content is computer-generated."
Fully AI, at volume. For Razorpay’s IPL 2026 season we made 32 founder films of 15 seconds each, directed by Shivang Monga, in one white world where each founder destroys the excuse they were told. Five days of set-up and a paid test film came first. After that the pipeline ran at about one film a day, and the 32 went out in batches over five weeks.
Hybrid, built around famous faces. In Malaysia Airlines’ Cricket at 30,000 Feet, Rohit Sharma, Hardik Pandya and Trent Boult turn a business class cabin into a pitch for one young fan; the 85-second film had passed 141 million views on Malaysia Airlines’ own YouTube channel by September 2026. It sits on our hybrid list beside Performax’s film with Jasprit Bumrah, Campa Cola’s with Ram Charan and Meta’s with Himesh Reshammiya. Every film on that list was shot first and extended with AI, so the generated parts inherit real light instead of inventing it.
| AI | Hybrid | Live action | |
|---|---|---|---|
| Cost in India, 2026 | From ₹21,000 to ₹75,000 on studio rate cards; about ₹4 to 12 lakh reported for brand-grade films | Set by the shoot days you keep | ₹3 lakh to ₹1.5 crore and above, by our tiers |
| Time | About a day of generation per short film once the stills are locked; for Razorpay’s 32, five days of set-up, then batches over five weeks | The shoot schedule, plus more post | Six to ten weeks from approved script to master |
| What you control | The world, the scale, the style | People and product on set; the world in post | Everything in front of the camera, nothing after the shoot without a reshoot |
| Best for | Fantasy, scale, impossible physics, many versions | Real people in impossible places; real products in made worlds | Performance, faces, food, celebrities, proof |
| Watch out for | Sameness, a mandatory label, faces that drift, changes after the lock | Matching light between two pipelines | The cost of a reshoot |
One script, three ways
Here is the decision made on a real page. In our guide to writing an ad film script we wrote Atlas, a 30-second sample film for a fictional atta brand, Suraj Atta. Meera, eight, rolls a roti that comes out shaped like Australia. Over the following weeks her father, Dev, keeps every country she makes in a steel dabba. At seventeen she finds the dabba. He is caught in the doorway. Pack shot.
Run the brief test first. The film has to prove something real: the rotis are the product’s result on screen. Two faces carry it, played by actors you will cast rather than famous people. It is one film. Everything points to a shoot, with AI at the edges. In our guide to the ad film script breakdown we read it as a producer would: one kitchen, two day parts, three principals, a child artiste on a five-hour day, a food styling job for six country-shaped rotis, and a complexity band of standard, the band where most national TVCs in India sit.
As live action, Atlas is exactly what it was written to be. Made as a broadcast national TVC, a standard-band script sits in the ₹25 to 60 lakh tier of our cost guide.
As hybrid, most of Atlas should still be shot. The film is two faces and a dabba. AI earns its place in four spots: the animatic that lets the client see the film before the pre-production meeting, the passing weeks outside the kitchen window, the end-frame environment around a real packshot, and the language versions. Two things should stay real. The second Meera should be a second actor: ageing a real eight-year-old’s face to seventeen with AI alters a real child’s likeness, which needs her guardian’s written consent and turns that footage into synthetically generated information that platforms must label. And the rotis should be real food from a real food stylist, because ASCI’s final guidelines prohibit exaggerated results.
Hybrid does not halve this film’s budget. It moves a little money from the shoot day into post and buys the client a clearer picture before anyone books a crew.
As fully AI, Atlas is possible and much cheaper on the production side. It is also a different film: a father and daughter across nine years whose faces have to hold, rotis that have to look like food, and a label on a film whose entire point is that a real father kept real rotis. Unless the film stops pretending. An animated Atlas, plainly drawn, is a different film and could be a good one, and it sits outside the IT Rules’ labelling duty, which covers AI footage that could pass for real people or events.

What do India’s AI ads so far tell you about the choice?
India’s first wave of AI ads already shows where the line falls. Audiences praised AI when it built something no camera could film, and criticised it when it stood in for something they expected to be real.
We gathered 17 Indian AI and hybrid ad films released or awarded between March 2025 and August 2026, each confirmed as AI-made in a brand release, the trade press or an awards entry, and recorded the reaction on record for each.
| Film | Route | What the AI made | Reaction on record |
|---|---|---|---|
| AJIO, The Great Fashion Price Crash, Mar 2025 | Fully AI | Two films, by EiPi Media | None found |
| Emami, DermiCool Warriors, May 2025 | Fully AI | Heat-rash monsters, in 30 days | None found |
| Tata Gluco+, Piyo Goodness, Karo Greatness, May 2025 | Fully AI | Every frame; no live action | 20 of the 50 most-liked YouTube comments attack the AI |
| TVS Jupiter 110, IPL 2025 | Fully AI | Characters from its original newspaper ad | Little on record |
| Danone Protinex, Jun 2025 | Fully AI | A narrative meant to look "as real as real" | None found |
| Siyaram’s, Papa Hain Toh, Jun 2025 | Fully AI | Visuals, voice and music | None found |
| CRED with Gautam Gambhir, Sep 2025 | Hybrid | An AI Gambhir, then the real one | afaqs: "nothing short of grotesque" |
| Swiggy Instamart with Ajay Devgn, Sep 2025 | AI inside a star film | Devgn surrounded by his own duplicates | Manifest: "bold, quirky, and memorable" |
| Tata Play, Samajhdar Bano, Oct 2025 | AI characters | A talking owl and donkey, in four films | Manifest: it "makes complex ideas feel simple" |
| Himalaya BabyCare, India Tiny Talks Summit, Oct 2025 | AI characters | Talking babies, in eight languages | Manifest: "more on cuteness than craft" |
2025.
| Film | Route | What the AI made | Reaction on record |
|---|---|---|---|
| Dinshaw’s rap battles, Feb 2026 | Fully AI | A rapping grandmother, stylised | None found |
| Sprite with Sharvari Wagh, Apr 2026 | AI versions around a real star | City-by-city versions | Praised by PRmoment; a Kolkata frame reportedly showed Sydney’s bridge |
| Bajaj Group centenary, directed by Rajkumar Hirani, May 2026 | Fully AI, 10 minutes | Gandhi and Jamnalal Bajaj | Praised for its storytelling; The AIDEM asked if Gandhi’s "approval" is a fake endorsement |
| Hyundai, 30 years in India, Jun 2026 | AI-led | Three decades of the brand’s past | Critics in afaqs; the most-liked YouTube comment attacks the AI |
| Tata Simply Better, #GoHexaneFree, Jun 2026 | AI-generated | Food that will not enter the oil | Prathap Suthan: it "undermines everything the brand has spent years saying" |
| Agoda India, Whattey Save (Peacock), awarded 2026 | Hybrid, shot first | A peacock and a dolphin in place of dummy props; an aged character | Good Ads Matter 2026 Silver |
| Motorola edge 70 max and moto g max, Aug 2026 | Fully AI | Sloths in a cinematic world | None found |
2026. Both tables: 17 films chosen by press coverage, not a sample. Reactions are those on record in the trade press and YouTube comments, not measures of sales.
Read down the last column and the brief test reappears. The praise went to films that put AI where no camera could go: a talking owl and donkey for Tata Play, a crowd of Ajay Devgns for Swiggy Instamart, a peacock where Agoda’s set had a dummy prop. The criticism went to films that asked AI to stand in for something viewers knew to be real: Hyundai’s own history, food going into Tata Simply Better’s oil, Gandhi’s endorsement in Bajaj’s centenary film, a Kolkata skyline, an AI version of Gautam Gambhir. The line is not about real people as such. A crowd of Devgns is plainly impossible; a stand-in for the real thing is not. None of this is a law, because 17 films picked by coverage cannot be one, but it is the same line the brief test draws.
We have since judged India's AI ads film by film, with the awards each won and what each got right or wrong, in our review of the best AI ads in India.
The comment sections tell the same story. We sorted the 50 most-liked YouTube comments on four films on 9 October 2026. On the two films announced as made with AI, Tata Gluco+'s and Hyundai’s, 27 of the top 100 comments attacked the AI, and the most-liked comment on each was a complaint about it: on Hyundai’s, that making the whole ad through AI ruined the feel to connect; on Tata Gluco+'s, sympathy for the actors and animators it replaced. On Air India’s AI film and Malaysia Airlines’ hybrid film, both built around cricketers, none of the top 100 mentioned AI. They talked about the players. Comment sections are a small, self-selected slice of any audience, but the split is hard to miss.

Brands abroad have learned the same thing faster and louder. McDonald’s Netherlands pulled its AI Christmas film three days after posting it in December 2025, after viewers called it "creepy" and "poorly edited", although its production company said the team had worked for seven weeks. At home, Cadbury Dairy Milk Silk, which had used generative AI for its 2024 Valentine’s campaign, answered in January 2026 with AI Knows Nothing, a live-action film by Ogilvy India and Good Morning Films. Its writer, Ogilvy’s Akshay Seth, told afaqs the point was that love should not be automated, not that technology is bad.
The testing data is kinder, until the word AI appears. Researchers at NYU Stern and Emory found that ads made entirely with generative AI drew click-through rates up to 19% higher than ads by human experts, and that telling people an ad was AI-made cut its click-through rate by 31.5%. Globally, a Harris Poll survey released at Cannes Lions in June 2026 found 73% of consumers would trust an ad less if they suspected AI had made it. Critics have moved on to sameness: Prathap Suthan of Bang In The Middle told afaqs in June 2026 that once anyone can generate spectacle, production value stops signalling that a brand cares.
The industry now judges the blend as a craft of its own: Good Ads Matter, which folded AI into a visual effects category in 2025, gave it a category of its own in 2026, with a separate award for AI-integrated live action.
What comes next is contested, and every prediction here is a forecast. Studio Blo’s Dipankar Mukherjee expects 90% of the world’s ad films to be made by AI rather than shot on camera by 2027. Martin Sorrell of S4 Capital set a four-month, $2.5 million shoot against four weeks and $500,000 in October 2025, then told The Drum at Cannes in June 2026 that "the answer, in the short to medium term, is human creativity." EY, cited by Reuters, projects AI could cut costs in Indian media and entertainment by 15% over the medium term. Ours is narrower. Ganesh said it to Social Samosa in April: "A camera can only do what a camera can do. But with AI, I can do things a camera never could." The brands that do well with AI will use it for exactly that, and keep the camera for everything the audience needs to be true.
Which rules now sit inside an AI ad budget in India?
Four sets of rules now apply to any AI or hybrid ad film made in India: platform labels under the IT Rules, ASCI’s new guidelines on AI in advertising, the courts’ orders on faces and voices, and the Copyright Office’s position on authorship. Each one is a line on the budget, and together they are why the first question of the brief test matters most.
- Platform labels. The IT Amendment Rules, 2026, in force since 20 February 2026, cover "synthetically generated information": AI-made or AI-altered audio and video that looks real and could pass for a real person or event. Platforms must label it prominently and embed metadata, and the largest ones, with 50 lakh or more Indian users, must ask the uploader to declare it. The brand uploading the film is that uploader. A draft amendment that would keep the label on screen for the whole film closed for comments on 7 May 2026 and has not been notified. On 8 October 2026 MeitY went further in an advisory to platforms, reported by MediaNama, asking them to label or add context to manipulated, synthetic or misleading content, particularly where it is presented as a current event.
- ASCI’s final guidelines. The Advertising Standards Council of India released its guidelines on labelling synthetically generated content in advertising on 29 September 2026; they take effect three months after publication. Banned even with a label: fabricated endorsements or testimonials, exaggerated results, non-existent locations shown as real, and anyone’s likeness without consent. Label required: synthetic influencers, a real person’s likeness or voice in personalised messages even with consent, fabricated events or settings that may change how consumers understand the product, and products that do not exist yet. No label needed: obvious fantasy, decorative backgrounds, routine edits. ASCI suggests wording such as "Video created using AI", says a label never cures a misleading ad, and reaches television through the Cable Television Networks Rules, which bar ads that break its code.
- Faces and voices. Indian courts keep stopping unauthorised AI use of a person’s identity: the Delhi High Court for Anil Kapoor in September 2023, and the Bombay High Court for Arijit Singh in July 2024 and for Asha Bhosle in September 2025, both over AI voice cloning. Budget for written, scoped consent and a fee for every real face or voice.
- Authorship. On 31 August 2026 India’s Registrar of Copyrights refused to register an AI system as an author, holding that the law recognises the human behind a work, not the tool. Keep a record of the human decisions, and settle who owns the finished film and the generated assets in the contract.
If the film is headed for awards, add one more. Cannes Lions requires every entry to say whether and how AI was used, added an AI craft subcategory to Film Craft for 2026, and can ban entrants who wilfully submit false or misleading work for up to three years. Document the AI work from the first day, not the week of entry.
In budget terms, that means four lines on every AI or hybrid quote: a labelled master and a clean master, a consent and licence file for every real person and every tool, a record of the human decisions behind the generated material, and legal review time before air.

When does this advice not apply?
Skip this if you are making performance creatives to test at volume. Variant ads for social testing are a different product, priced per creative or on retainer, and a production house is the wrong supplier for them.
Skip it if the film is built around a celebrity. Their fee, their calendar and their consent set the cost, and AI rarely makes that cheaper.
Skip it if your category makes claims about health, money or results. ASCI’s final guidelines prohibit fabricated testimonials and exaggerated results even with a label, and generated proof is the fastest route to both. Shoot the proof.
And skip it if you need the film on air this week but have not decided what it should be. AI is fast once the creative decisions are made. It is not faster at making them.
How do you brief an AI or hybrid film so the quotes compare?
Our cost guide lists six things to send every production house: the script and runtime, every deliverable, the talent expectation, usage rights, the air date and your budget. For an AI or hybrid film, send those and four more.
- The shots that must be real, and what has to stay consistent. Faces, product, proof. Mark them on the script.
- Every format and language. Video models set the frame shape when each clip is generated (Google’s own model makes horizontal video unless 9:16 is asked for), so a vertical version is often a separate generation rather than a crop. Name 9:16 at the start.
- Every real person involved, and the consent you already hold for their face or voice.
- Masters, ownership and tools. Ask for a labelled and a clean master, say in writing who owns the finished film and the generated assets, and ask which models the film will be made on. The tools move fast: OpenAI removed Sora’s API on 24 September 2026, and of the main video models only Google, for Veo on its paid cloud platform, and Adobe, for its own Firefly video model on enterprise contracts, offer IP indemnity.
Then ask each bidder two questions. At which stages are revisions included, and what resets the clock? The answers tell you more about the quote than the number does.

Get a producer’s read on your brief
Every brief already knows which route it wants. What the film must prove, who must be on screen, how many films it is and how fast you can sign off are all settled before anyone quotes.
If one scene feels too risky to promise, bring it to us as a one-scene feasibility test before full production. You can also paste your script into our script analyser for a free producer’s read in about a minute, or send us the brief and a producer will tell you which shots to shoot, which to generate and which to blend.
First December Films is an advertising film production company in Mumbai. Our work has won more than 150 international creative awards, including Cannes Lions, D&AD and One Show Gold Pencils.
Frequently asked questions
How much does an AI ad film cost in India?
Studio rate cards in India start a 30-second AI ad film at 21,000 to 75,000 rupees, with a median near 45,000, as of October 2026. AI films made to a brand’s television standard are reported in the trade press at roughly 4 to 12 lakh rupees, against about 25 lakh rupees for a broadcast live-action TVC.
Is an AI ad film cheaper than shooting a TVC?
Usually, but by less than the headlines say. Indian trade examples put a brand-grade AI film at roughly 13 to 25 percent of the shoot or CG film it replaced. The saving shrinks when the film depends on a celebrity, real locations or a full edit, grade and sound pass, because AI removes none of those.
How do you decide between AI, live action and hybrid for a brief?
Answer four questions before the script: what the film must prove, who must be on screen, how many films and formats it needs, and how early each scene can be signed off. Shoot proof and famous faces, generate invented worlds and volume, and test any scene that will not settle before committing the full budget.
What is a hybrid ad film?
A hybrid ad film is shot first and extended with AI. The people, the product and the proof are filmed on a real set, then the world around them is generated, such as set extensions, backgrounds, crowds or metaphor shots, so the generated parts inherit real light. Its cost is set by the shoot days it keeps, plus compositing in post.
Do AI ads need a label in India?
Often, yes. Since 20 February 2026, India’s IT Rules require platforms to label realistic AI-made video and audio of people or events. ASCI’s final guidelines, released on 29 September 2026, require labels for uses such as synthetic influencers and fabricated settings, and ban fake testimonials and exaggerated results outright. Obvious fantasy and routine edits need no label.
How long does it take to make an AI ad film?
Once its characters, costumes and locations are approved as stills, a 15 to 30 second AI film takes about a day of generation in our pipeline, and our 32 IPL films for Razorpay went out over five weeks after five days of set-up. Published Indian examples run from four days for AJIO’s 2025 campaign to 30 days for Emami’s. A live-action national TVC takes six to ten weeks.
Who owns an AI-generated ad film in India?
The contract decides it. On 31 August 2026 India’s Registrar of Copyrights refused to register an AI system as an author, holding that the law recognises the human behind a work, not the tool. Agree in writing who owns the finished film and the generated assets, and keep a record of the human creative decisions.



