Inspiration
We've played RPGs before. We love RPGs, but sometimes, the dialog options can take away from immersion. Sometimes, options can force you to pick something you don't want to pick, and there's usually no way to let game characters down easy. Sometimes, you get attached to video game characters and have no way to pursue more than just the dialog options. This would change all of that.
What it does
Our colab notebook lets you train your own generative text model and play around with it for yourself! For the purposes of time, we made our dataset small and the amount of training abysmal, however, it is proof that this approach could work, and that a conversational engine could be built with what we have here as a framework.
How we built it
Google Colab. Lots, and lots of Google Colab. We used a dataset from a human pretending to be a bot on some instant messenger apps as our data (dataset can be found here: link) to train a machine learning model to generate responses to our queries. The model was made with Marvelous designer.
Challenges we ran into
Firstly, neither of us had any experience doing any of what we did. There was a lot of googling, and a lot of research that went into this on the first day of the hackathon. Next, it takes a long time to train ML models. Colab also does not store models for you, as we found out the hard way. We trained no less than three models, each of which took two hours.
Accomplishments that we're proud of
We got through knowing nothing, and now we're better versed in the fields that we put our work toward this hackathon!
What we learned
We learned more about Machine Learning, 3D art, and what it takes to eventually release a game!
What next?
This is the initial steps taken toward playing with game design. We plan on going forth and releasing our work as a standalone game when we are done!
Log in or sign up for Devpost to join the conversation.