The original question was closed for being too vague, but can be found here for those interested in details.

Basically, if I have a model like this or this that is already built for speech-to-text capabilities, but I only want it to choose from a specific set of words, or add new words, how would I modify/re-train/fine tune the existing model?

Based on research, it looks like I might need to gather my own data and training examples, which is no problem. I just want to make sure I’m on the right track. I have verified that the model works using a basic Python script, but I don’t know how to mess with the internal structure or repurpose it.