I wanted to see what the buzz with machine learning and AI is. So I proceeded to use two separate libraries and learn none of those. Because the libraries encapsulate all the actual hard concepts.
This program will use your camera and draw super saiyan hair for you.
But you also need cheesy, long-winded monologues to really stick it to the villain (or hero). So it will generate you some DBZ dialogue for you to scream out, you method actor, you.
markovify is a library that takes in text (called a corpus), find patterns (markov chain), and generate new, emergent text (text generation).
face_recognition is a library that has machine learning models that have been trained with faces.
It's scarily accurate, with only one photo as input data.
This repo isn't turnkey. For good reasons. You need text and I don't want to distribute any copyrighted material. You need photos and I don't want to give you mine or my friends.
The markov chain requires data, which is up to you to find. I recommend finding some subtitles of Dragon Ball Z.
Put the files into sources/sanitized/.
All files in this dir will be fed in a single, common source of text.
Use the same anime, if you want consistent results.
Or maybe it doesn't matter because most anime is deriviative (braces self with flame shield).
Put .jpeg photos in face-data/.
The name of the file, sans extension, is the name of the person.
python facerec_from_webcam_faster.pyWill start the camera and face detections.
python markove_me.pyWill read corpus and generate random sentences. Hit to continually generate new lines.