Sarvam Vision

Hey there, future tech explorers! Welcome back to our “Sovereign AI” journey. 🚀

In our last post, we met Sarvam AI—India’s own super-digital-brain. Today, we are opening one of its coolest tools: Sarvam Vision. Think of this as the “Super Eyes” for AI. It doesn’t just look at pictures; it actually understands them, even if they are written in Indian languages!

Let’s dive into the lab and see what these magic eyes can do.

👀 What is Sarvam Vision?

Imagine taking a photo of a messy handwritten note or a sign at a train station in a language you don’t know perfectly. Normally, your phone just sees a bunch of pixels. But Sarvam Vision reads the words, understands the tables, and can even describe the whole scene to you in a story!

This is the “Control Center” where you upload your photos to give the AI its super-sight!

🛠️ The 4 Superpowers of Sarvam Vision

Based on our experiments in the lab, here are the four main things you can do:

1. The Handwriting Decoder ✍️

Have you ever looked at a teacher’s messy notes or an old recipe and thought, “I can’t read this!”? Sarvam can.

• What it does: It looks at handwriting (like Hindi proverbs on a notebook page) and turns it into clean, typed text.
• Example: We gave it a page of Hindi proverbs written by hand, and it typed them out perfectly on the screen!

2. The Magic Table-Maker 📊

Sometimes information is stuck inside a photo, like a schedule on a TV screen or a menu. It’s a pain to type it all out.

• What it does: It finds rows and columns in a photo and turns them into a tidy table you can actually use.
• Example: We showed it a photo of an Airport Arrival Board written in Kannada. Sarvam didn’t just read it; it organized the flight numbers, times, and cities into a beautiful list!

3. The AI Storyteller (Image Captioning) 🖼️

This is like having a friend describe exactly what’s happening in a photo.

• What it does: It looks at a scene and writes a detailed paragraph about what it sees.
• Example: We uploaded a photo of a Railway Station. Sarvam wrote a long description in Hindi, explaining that there were two tracks, people waiting on the platform, and a yellow safety line. It’s like the AI is actually standing there!

4. The Sign Reader (Read Text) 🪧

Street signs and notice boards can be hard to read if they are far away or in a different script.

• What it does: It zooms in and pulls the text off any sign.
• Example: We tried a big black notice board in Gujarati. Sarvam read the whole thing and gave us the text to copy and paste.

⚡ Turbo vs. Pro: Which one do I pick?

When you use Vision, you’ll see two buttons: Turbo and Pro.

• 🚀 Turbo: Use this when you want an answer super fast (like reading a quick sign).
• 💎 Pro: Use this when the picture is really hard to read or you need the AI to be extra smart (like reading very messy handwriting).

💡 When should I use these?

🌈 Why this is so cool for India

Most AIs from other countries struggle with Indian languages like Kannada or Gujarati, especially when they are handwritten. Sarvam Vision is a local hero because it was trained specifically on Indian signs, Indian handwriting, and Indian scenes. It’s AI that finally understands our home! 🇮🇳

What’s Next?

Now that we’ve seen how Sarvam sees, in my next blog, we are going to learn how it speaks! We’ll be checking out the Text-to-Speech tools to see if we can make the AI sound like a real person.

What would you want Sarvam Vision to “look at” for you? Tell me in the comments! 👇

Note: All experiments were performed on the Sarvam AI playground. Stay tuned for more!