Build me a tool that captures data from a stream similar to the one linked below

Hi please include in your message any openCV / image/videostream processing, OCR experience you may have, I've seen c++ mentioned as a performant option, but the stack is not centrally important, just confidence and speed in the task. I'm looking for help developing a tool that identifies when an item appears in a [login to view URL] stream, and captures the text which appears above it. I've included a few examples to serve as proof of concept. One is called rage-analytics which I've linked below, but instead of capturing faces and sentiment analysis, the task would be to identify when certain objects in [login to view URL] streams appear, along with the text which appears with it, this occurs in two distinct patterns within the frames of the stream. Your task would be to capture text from a stream when it appears and mark the date and time it was captured. The data should ideally be put into CSV files to make this iteration faster, in the future more work is possible to move them into a database or manipulate further. [login to view URL] [login to view URL] If you'd like to submit a proposal for a version of this tool which only targets past videos, and not live streams, I'd consider accepting it if it would be a simpler/faster implementation. I've included the video API docs below [login to view URL] I've included another project where the relevant piece being where it captures playing cards and game scores from the Twitch stream. [login to view URL]