Sounds like the original voice - really great work.
11 Labs also does mouth‑syncing and they've been at it for ages; this is just the beginning for Sarvam, which is nice to hear.
Impressive, if it's real.
He's an excellent teacher and one of the most sought‑after faculty at IIT‑Madras. Students were thrilled when he became the director. He managed to deliver Shakti despite the usual Indian bureaucracy. IITs and IIMs used to have their own ecosystem until cylinder ma'am decided to curb their autonomy. Now the selection of directors and HoDs seems to depend more on political leanings than on academic merit.
So you guys haven't actually used Sarvam, huh? All these assumptions are just from an Instagram reel?
Beaten what exactly? I tried their OCR model - they claim it beats other SOTA models, but I still found Gemini/Chandra to be better.
Does Sarvam have its own foundational model, though?
Wow, this is good!
It's already existing tech, but okay.
I was there when they showcased it at the summit - very cool! They also did an ISRO demo.