Our Amar Cotton College music video began as a shoebox of faded photographs from 2001 to 2003. Each one went through five stages: the old photograph, the empty place, the paper people, the film frame and the motion. It took 104 location plates, 116 storyboard frames and 97 video clips, and not one real face. Here is exactly how I made it, with the numbers, the mistakes and the lessons you can use.
The making-of film, in my own voice, ending with eight of our photographs taken through all five stages. The music video itself is here: Amar Cotton College.
1. It started with a song
Before any picture, there was the song: four minutes and forty-six seconds of Assamese lyrics about the river, the old trees, the canteen, our friends, and a promise that distance cannot erase. Every shot was later designed on a sung line, so the song decided the film, never the other way round.
The brief I set myself was simple to say and hard to do. Make every viewer feel their own college years. Show the real campus. And never put a real person’s face on screen.
2. Three layers of time
I gathered everything I had: frames from an old film about the college from the 1970s, our own batch photographs from 2001 to 2003, and new photographs of the campus as it stands today. That gave me three layers of time. The 1970s as our legacy. Our own years as memory. Today as the present. The whole film moves between these three.
3. Clean the place before you fill it
Step one was the places. Every old photograph became a clean, modern photographic plate of the same place: the people removed, the architecture kept exactly as it was. The centenary arch in 2001 keeps its carvings, its lettering and its trees; it is only sharper, brighter and empty, so the story can walk into it. Our chemistry laboratory keeps its bottles, its counter and its blackboard, with my classmates gone.
Fifty-two places, two versions of each: 104 location plates, every one logged with the exact instruction that made it.
Tip: describe exactly what must stay, the timber, the windows, the light, and ask only for the people to go. Never ask for a new building.
4. One idea holds the film together
In Assam the xewali, the night jasmine, flowers at night and falls at dawn. In this film a falling xewali is the door into every memory. Whenever a blossom falls, we move in time: a student from the 1970s walks into the light, and the flower lands in today’s empty corridor. The clearest example is one porch, one camera: the 1970s, then our batch in 2001, then today, each era arriving on a falling flower.
5. Paper people in real places
For the people, I chose papercraft: folded paper figures standing in real, photographic places. They feel like us, but they are nobody’s exact face. The paper itself tells you the year: aged ivory card for the 1970s, bright craft paper for 2001, heavier card with silver at the temples for 2026.
Each character was built from our real clothes, taken from the batch photographs: check shirts, a film camera, a satchel. Then I aged them. The boy with the camera becomes the man who comes back with a shoebox of photographs; the girl with the poem becomes the woman flying home. Then came the crowds (Cottonians of the 1970s, today’s alumni, our teachers) and the props, like the 2002 college magazine with a pressed flower inside. Twenty character and prop sheets in all.
6. How one frame is built
One frame is four pictures and one written scene: a location plate, a character sheet, a prop and the flower. The written part matters most. Where we are. Who is here. What happens. Where the camera stands. How the paper looks. The light of that era. Every instruction is complete on its own.
Tip: always name a real lens and a real camera height. That one habit turns a flat picture into a frame that feels photographed.
Forty-seven shots became fifty keyframes, one for every line of the song, each checked by eye before it was kept. For many shots I made a second frame, the next moment from a new angle, with the first frame attached, so the people, the clothes and the light stay the same.
7. Fixing every mistake, and keeping the record
Not everything worked the first time. The shoebox came back with the wrong people on its top photograph. In one frame the woman appeared twice. In another, the flower grew to the size of a football. Every fault got a corrected version, and the old one stays in the record. Every image is logged: the file, the exact instruction, the references it used and where it lives on the disk, 240 records by the time the keyframes were done.
8. The moments
Last came the moments that make a music video stay with you, written from memory and placed between the lines of the song. The man sitting beside the boy he used to be, on the bench they once shared. One earphone each, by the tank. Fritters in the rain. A gamosa placed around a friend’s neck on the last day. A snowfall of torn notes from every corridor. Ten small moments in all.
9. From storyboard to motion
Everything came together in one storyboard: 116 frames in the order of the song, each timed to its sung line. Each frame then became a short clip, 97 in all. Where the camera never cuts, a clip gets a first frame and a last frame and moves between them in one unbroken take.
The motion instructions stay short: only what moves, and one camera move. The picture already knows what everything looks like. Exaggerated motion makes paper people behave strangely, so each clip carries one small human moment, a laugh, a glance, a hand on a shoulder. Natural sound only; the song goes on top in the edit.
10. What I did not do: change anyone’s face
Our batch photographs are of real friends, so I was careful about what AI touched. I tried restoring some of them for the end credits. The colour improved, but the faces changed, and a photograph of a friend with a different face is not a photograph of that friend. So I stopped. The end credits show the original photographs as they are, and in the music video every person is papercraft. The real photographs appear only in this making-of, beside the frames they became, with my classmates told first.
Five lessons
- Clean the place before you fill it.
- Say what must stay, not only what must change.
- Name the lens and the camera height.
- Carry continuity forward: attach the last approved frame to the next one.
- Keep every version, and choose only at the end.
Why AI tools alone cannot make a film
The tools will keep changing every few months. The order of the work does not: first the song, then the places, the people, the frames, and only then motion. Knowing which memory belongs on which line, where a camera should stand in a 1970s corridor, or that a xewali falls at dawn, comes from two decades of making films with real actors, cameras and locations, and from having walked under that arch myself. AI made the pictures move; direction decided what they meant.
The eight journeys
The film ends with eight of our photographs taken through all five stages, start to finish, with the song: the canteen, the chemistry laboratory, the centenary arch, the Dighalipukhuri railing, the old staircase, the reading room, the practical classroom and the founder’s bust.
Have old photographs that deserve a film?
The same pipeline turns the photographs of schools, colleges, institutions and families into anniversary, exhibition and memorial films. See our archive and restoration films or photo restoration, and tell us about your project.
More from the studio: the music video and its lyrics, music and music videos, and the making of Maa Kali vs Raktabeej, the making of Radha Krishna and the making of The Eternal Flute.
Lyrics, music, direction and animation: Tarunabh Dutta · assisted by Himakhi Dutta · sung by Jishnu Baruah and Jyotismita Baishya · a TD Film Studio production.
Frequently asked questions
How do you turn an old photo into a video with AI?
Not in one step. We rebuilt each place as a clean, empty photographic plate, designed the characters separately, wrote one scene per frame to combine them, and only then animated the frame with one camera move and one small action.
Can AI animate old photos without changing the faces?
Not reliably. When we tried restoring our batch photographs with AI, the colour improved but the faces changed. So no real face is in the music video: the people are papercraft, and the original photographs appear only as they are.
How many images and clips were made?
104 location plates, 20 character and prop sheets, 116 storyboard frames and 97 video clips, with every image logged against the exact instruction that made it.
Is the music video AI-generated?
The pictures and motion were made with AI animation tools under human direction. The song, the story, the characters, the storyboard and the edit were made by people, and the video carries YouTube’s altered-content label.
Can you make a film from my old photographs?
Yes. Institutions, schools, colleges and families send us their photographs, and we turn them into anniversary, exhibition and memorial films. Start with the archive films service.