Create ML: Create Text Recognition Model?

I am a newbie to the machine learning, so sorry if my question sounds naive...

I've been reading Vision's API to see if it is possible to recognize text, and realize that currently Vision only recognizes text's "rectangle" via VNDetectTextRectanglesRequest. After watching WWDC 2018's video about Create ML, I start to wonder if it is possible to use Create ML to create a custom model for recognizing the text (or more specifically, characters). My imagined workflow would be:

1. Using Vision to detect the bounding box of characters in an image
2. Cropping the character images and labeling them
3. On the Mac, using Create ML to train the model

Although I have a gut feeling that the framework is for reconizing the real world objects, not for the texts...