What it does
Enter in two Twitter handles and get back a generated tweet trained on the combined corpuses (corpi?) of real tweets from those users.
How we built it
The site was written in flask and the backend used Python. For the actual tech, we use the twitterscraper API to fetch past tweets from users. After combining the two bodies of text, we used a probabilistic model that looked at the k past characters to generate the next character, and used the corpus of text to sample from the possible next characters. We seed using a real substring of k characters from the tweets.