Typing is hard, one can see this when we type people and there are many different types and there's rarely consensus.
VI typing alone is even more hard, there are simply way too many possible faces. To automate this, one would need thousands, if not dozens of thousands of photos per type to train AI. And there's problem, how to accurately type dozens of thousands of people in first place? This is why nobody has automated it yet, not even tried.
However if typing based on hand would be easier that would change the situation. Hand is a lot simpler picture and there are a lot less characteristics in it to analyse. Therefore for hand based VI there would need to be at least hundreds of photos per type to train AI. Assuming that typing from the hand is the real thing. I never researched this subject, but once I've spotted two SEIs with same hand, so similar that I would say it should be possible to VI them based on that. The hand in question was very slim and elongated compared to mine.
As example, I am posting my hand, taken by the webcam but you can see the whole hand. It's ILE hand, I wonder if there's any ILE who can confirm similar characteristics of the hand. Maybe you can post other hands for comparison, along with the type.
Also regarding automatic VI, one could focus on the eyes but that's also very difficult as eyes are always small in the picture and have variations like tired eyes, happy eyes etc. Hand on the other hand is big and clear so that should be really easier, if possible.