Case study
One lakh personalized videos in 20 days: how we did it for Syngenta India
One shoot. Ten languages. Twelve states. More than a hundred thousand films, each greeting the viewer by name, in about twenty days. Here is how it was built.
For the launch of Miravis Duo, Syngenta India wanted its Managing Director to personally greet every retailer, farmer and stakeholder by name, in their own language. That is not a message you can shoot. It is a system you have to build. This is the system.
The brief, and why the usual playbook would not do
The fungicide market in India has more than two hundred players, and most of them run the same launch: a generic film, one message, and hope. Syngenta India wanted the opposite. They wanted the launch of Miravis Duo to feel like a personal introduction from the top of the company to every single person who mattered to it: retailers, farmers, and stakeholders inside and outside the business.
The obvious way to do that is to shoot the Managing Director once and add a name caption. It does not work. A name in a text overlay is decoration. A name spoken aloud, in the language the viewer thinks in, is recognition. The brief was the second thing, at a volume no studio could staff.
What we built
- One master shoot. We filmed the Managing Director once, with a script designed so that names, places and language could be varied without breaking the performance.
- AI voice cloning, with consent and control. The MD's voice was cloned so that each recipient's name could be spoken naturally, in the right language, in his own voice. Every generated line was reviewed against the original.
- Avatar generation for lip-sync. The picture had to match the sound. Avatar generation kept the MD's delivery intact while the words changed from film to film.
- Ten Indian languages, twelve states. Native speakers reviewed the scripts and the outputs in each language, so a retailer in Tamil Nadu heard the MD in Tamil and a farmer in Punjab heard him in Punjabi, with their own name in the sentence.
- A clean data pipeline. More than a hundred thousand records had to become more than a hundred thousand correct films. Names, languages and delivery details were cleaned, validated and matched before a single frame was rendered.
- WhatsApp delivery. The films went straight to the phone, because that is where these audiences actually live. Nothing was made for a channel the recipients do not use.
The numbers
More than one lakh unique films. Ten languages. Twelve states. One shoot. From brief to final delivery, about twenty days. And the part that mattered to Syngenta: a measurable shift in engagement and brand recall, in a category where most communication is ignored.
Nobody in Indian agriculture had done this before, not at this scale and not with this level of craft. It is also the kind of implementation that earned Hoopla its place as an OpenAI Select Partner, because this is what applied AI looks like when a creative team actually builds with it.
What made it work, and what would have broken it
Three decisions carried the programme. First, the personalization was not cosmetic. The name was spoken, not overlaid, and the language was the recipient's, not a subtitle. Second, humans held the standard at every stage: the script, the master performance, the language review and the final QA were all people. The machine multiplied; it did not decide. Third, the data was treated as production material. A single bad record becomes a film that greets someone by the wrong name, so the pipeline was built to catch that before rendering, not after sending.
What would have broken it: rushing the master shoot, skipping native review to save a day, or trusting the data as delivered. Every one of those saves time on paper and costs the whole programme in practice.
If you are considering something like this
Personalized video at this scale is a system, and the first question is always whether the business case supports one: a launch with a large, known audience, a message worth delivering one to one, and a channel the audience already uses. If that is your situation, the shape of the programme is well understood now, and we are glad to walk through it. Start a conversation and tell us about the launch; we will tell you honestly whether it fits. The broader method is written up in how we make 100,000 personalized videos.
Questions, answered
How were one lakh personalized videos made from a single shoot?
The Managing Director was filmed once with a script built for variation. AI voice cloning and avatar generation then produced each film with the recipient's name spoken in their own language, while native speakers reviewed scripts and outputs in every language and a validated data pipeline drove the rendering.
Which languages and regions did the Syngenta programme cover?
Ten Indian languages across twelve states, with each recipient receiving the film in their own language, delivered on WhatsApp.
How long did the programme take?
From brief to final delivery, about twenty days, including the master shoot, language review, rendering of more than one lakh films and WhatsApp delivery.
What does a personalized video programme like this cost?
It depends on audience size, languages, channels and how much master creative is needed, so we do not publish a figure. Tell us about your launch and we will scope it honestly: hoopla.agency/contact.