Case Study

Voice assistants are all the rage, they are omnipresent from a 700 rupees Jio phone to Honda City (another blog on this coming up). From using it for breaking literacy barriers to operating your car without lifting a finger, Voice does it all.

But not every use case has to be so profound.

Here’s an imagination that came true. What if India’s most loved, timeless game was to be imagined in an AI and Voice First world?

We did exactly that and ended up with Gaana Antakshari on Google Assistant.

If you are reading this on your Android Phone, then just long-press your Home button, once Google Assistant comes to life, say “Ok Google, play Gaana Antakshari”, nuff said!

Ok Google, play Gaana Antakshari

If you are still here, here comes a more geeky piece of information on how we built this.

Firstly, this is what it does: You can sing a Bollywood song, the Assistant listens, finds the last syllabic and plays a Bollywood song back, and then your turn. Have fun, repeat :)

How did we build this?

Traditionally, (already using that word in Voice!) Voice Assistants are built based on understanding users’ ‘intent’. In this case, the intent is standard, you want to play Antakshari. So we rather used Scenario-based programming.

Once that was sorted, we had to ensure a wide library of songs is available for the bot to match the human intelligence level, so far we have about 900 Bollywood songs.

We then had to train the bot to understand two things, are you really singing a song or making stuff up, and next, understand & transliterate in near real-time the lyrics of the song a human is singing.

Just one more layer of complexity to solve, your Assistant language may be in English, but you are singing a Hindi song, so bot ko thoda waise train karna pada jaise average Indian bat karte hain, i.e. Speak more than one language at a time.

Now do this all in under 2 seconds, so the game is smooth and seamless, and more importantly, is fun to play.

Core tech ingredients.

Just like how Bollywood songs have many pieces like Lyrics, Vocal, Music coming together to create harmony, we had all the tech mix. NLP, ASR, NLU and Applied AI for making the Antakshari experience happen.

We have to thank do bade log for this innovation to come to life. Ek hai Gaana aur doosra, Google.

Aur aapke attention aur patronage ke liye, thank you very much from India’s first Voice Agency.

Are you thinking of using AI for such innovations? We are just an email away