Language
Size

English (Master cut) · 16:9 · 1920×1080 · 0:45YouTube · CTV · Website

Case Study — AI · Performance · Campaigns

Vedantu

Class 11 · 2-Year Batch — AI Performance Ad

One 45-second performance ad for Vedantu Learning Centres' Class 11 two-year JEE/NEET batch, written and produced with AI. From a single script it shipped as 16 finished videos: English, Tamil, and two Telugu versions (Andhra Pradesh and Telangana), each in four sizes.

कKaaran

Class 11 is when JEE/NEET preparation quietly slips. The ad had to meet that moment first (half the year gone, the plan not working) and only then make the offer: you're not alone, and there's still time.

खKhaaka

One emotion-tagged script. AI-generated students in the first half, real Vedantu Learning Centre footage in the second, kinetic on-screen text, and a region-specific end card. Adapted for Tamil Nadu, Andhra Pradesh and Telangana, in 16:9, 9:16, 1:1 and 4:5.

गGoonj

16 finished videos from one script: four versions in four sizes each, cut from the same AI shots and real footage.

The brief

A performance ad for Vedantu Learning Centres' Class 11 2-Year Batch (JEE / NEET): one master film, built to ship in English, Tamil and Telugu and in every size a media buyer asks for.

The film is a hybrid by design. The emotional first half (a student alone with a test paper at night) is AI-generated, so it can be re-shot and re-cast quickly. The second half is Vedantu's own Learning Centre footage, so the proof is real.

By the numbers

16finished videos from one script
4versions: English, Tamil, Telugu (AP), Telugu (TS)
4sizes each: 16:9, 9:16, 1:1, 4:5
45–48 sper video

The script, beat by beat

TimeOn screenVoice-overPicture
0:01–0:08Class 11. Half the Year… Gone. / JEE / NEET Prep Not Going to Plan?Half of Class 11 is over and the plan isn't working. (thoughtful, then worried)AI: a student studying late at night; a test paper close-up
0:10–0:13You're Not Alone.Plenty of students are in exactly this spot. (reassuring)AI: a tired student, hand on her cheek, over her books
0:13–0:16There's Still Time to Get Back on Track.The turn: there is still time. (hopeful)AI: morning light, the cold grade lifts
0:16–0:20India's Top JEE & NEET TeachersVedantu Learning Centres' teachers. (confident)Real: Learning Centre lobby and classroom
0:20–0:26Strong Basics · Doubt Solving · Boards + JEE/NEET, TogetherStrong basics, doubts solved, Boards and entrances prepared for together.Real: classroom footage
0:26–0:33Hi-Tech Classrooms · LIVE Quizzes · Smart Clickers · Dedicated Academic MentorWhat the centre adds to the classroom. (excited)Real: classroom and mentor footage
0:35–0:39Start Your JEE | NEET Prep — 2-Year Batch — Vedantu Learning CentresStart with the 2-Year Batch. (energetic)Offer card over a phone in hand
0:39–0:45Admissions Open · Take MVSAT, unlock up to 100% scholarship · Book a Free DemoVisit your nearest Learning Centre.Region-specific end card

Times are for the English master; the Tamil and Telugu versions run 2–3 seconds longer to let the voice-over breathe.

Language adapts

Tamil and Telugu were adapted, not just translated. ChatGPT produced the first drafts; Gemini reviewed and tightened them. The tightened versions ran about 30% shorter than the literal drafts and read as conversation. English words students actually say (JEE, NEET, doubts, classroom) stay in English, in Latin script.

Each language got its own voice, cast on ElevenLabs v3 with emotion tags, and the regional cuts got a new, longer music bed. No new video was generated: the same AI shots and real footage were re-timed to each voice-over.

English
Master cut. Half of Class 11 is gone and the plan isn't working: the film opens on that.
Tamil · Tamil Nadu
“Class 11-ல பாதி வருஷம் முடிஞ்சும், உங்க JEE NEET preparation plan பண்ணபடி போகலையா?”
Telugu · AP and TS
“Class 11 లో సగం year అయిపోయినా, మీ JEE NEET preparation అనుకున్నట్టు జరగట్లేదా?” One Telugu voice-over serves both states.
End cards
A separate CTA end card for each of Tamil Nadu, Andhra Pradesh and Telangana, built from supplied artwork.

Size adapts

Every language ships in four sizes: 16:9 (1920×1080), 9:16 (1080×1920), 1:1 (1080×1080) and 4:5 (1080×1350). The AI video model keeps the shape of its start frame, so the key shots were generated in both 3:2 and 2:3. The 3:2 clips feed the 16:9 and 1:1 cuts, the 2:3 clips feed 9:16 and 4:5.

On-screen text, labels and end cards are re-laid out for each size and rendered as transparent overlays, so a fix to one line of copy re-renders all 16 videos.

How it was made

  1. Brief and script. Vedantu's brand and content material plus public sources; a seven-beat script with an emotion tag on each beat, ready for ElevenLabs v3.
  2. Frames. The AI students were designed as still frames in ChatGPT: realistic, mild emotion, Indian faces.
  3. Motion. Each frame was animated with Kling 3.0 on Higgsfield at 1080p. Six clips for the first cut.
  4. Real footage. Vedantu Learning Centre B-roll cut in for the proof: lobby, classrooms, mentors.
  5. Voice, music and sound. ElevenLabs v3 voice-over, a music bed, and sound effects on the on-screen beats. Music is ducked under the voice.
  6. Motion graphics. Supers, pills, labels and the offer card are HTML and GSAP, rendered through headless Chrome to transparent ProRes, then composited and mixed with ffmpeg.
  7. Client feedback. Round 1 came back; six shots were re-generated from new frames and the English cut revised.
  8. Regional versions and sizes. Tamil, Telugu (AP) and Telugu (TS) re-timed to new voice-overs, each rendered in four sizes. Twelve more videos, no new video generation.
Process Artefact
Revision note, 25 Sep 2026. After client feedback round 1, six shots were re-generated from new student frames and the sound effects were refreshed. The regional versions followed the same day.
Outcome
16 finished videos from one script · 4 versions × 4 sizes