How we built a vernacular video captions AI

TanglishCaptions was not born out of a big startup idea.

It was born out of pure pain and a lot of laziness.

The Problem

During one of our tech community meetups, we recorded a lot of testimonials. Like a lot lot.

Phones out, cameras rolling, people sharing honest thoughts. Everything felt great.

We even had a paid subscription for a popular AI captioning tool. So we confidently shot more videos thinking editing would be easy.

Then editing day came. And that is when reality hit.

Everyone spoke in Tamil. But the AI didn't.

English captions was possible. Tanglish captions was impossible.

Manually captioning everything was not happening. We are lazy to the core.

Instead of spending 4 hours to caption manually, we spent 40 days over engineering it, and saw outcomes that were too good to be true.

We had no idea how to build language tech solution but we learnt, unlearnt and over engineered it.

We knew we are not the only ones with this problem.

We had Sanjeev N C (Founder of Supermeme.ai). As a product studio, we worked end to end with Sanjeev N C.

From idea to MVP. From design to tech. Sanjeev led the launch and marketing. We handled the product and engineering side.

The goal was simple:

What earlier took nearly 4 hours of manual effort now takes under 40 seconds. That alone felt good.

Under the hood it uses popular AI models and Deep Learning models to achieve not textbook Tamil, but normal spoken Tamil.

The hard part was not transcription. It was getting the feel right.

Tanglish is emotional. It is messy. It is very human. And that is what we tried to preserve.

Right now, TanglishCaptions is in beta. Breaking things. Fixing things. Breaking them again.

If you are interested to know more about this product, let us know and we would be happy to set up a demo for you.

Remember: Ideas are cheap, Execution is everything.