Delivery Run

Icon

Inspiration

Top down retro games that you would play on a game boy like Pokémon Emerald with updated graphics. We also wanted to incorporate generative AI like popular games infinite craft where they use AI to make an infinite amount of possibilities in game. We also wanted to tie everything back to the sponsor Amazon by building a custom web extension that ties directly to the Amazon website.

What it does

Play in a world where Grandma's cats have gone missing and your job is to find them. Talk to NPC's that use Generative AI to interact with the player and buy items on Amazon to help them out along the way.

How we built it

We started with an ambitious idea and stripped it down to it's core mechanics and used Unity3D and Visual Studio 2020 to make it possible:

Dialogue We built a custom dialogue editor from scratch to be able to branch into different dialogue options. This modular system allowed us to easily make more and more NPC's without compromising time. The Dialogue editor allowed us to use AI responses with the click of a button whenever it was necessary.

Web Extension It was the first time we had ever built a web extension and actually have it work on the Amazon store. The extension made it possible to use all of Amazon's amazing features without having to rebuild a much worse-copy in game. The extension began very complex and eventually we simplified it. The mechanics of the extension finds a product using the HTML 'Product Name:' and the 'Product Price' and copies it on the user's clipboard. Then it is decoded and parsed into a struct that can be used for shopping, equipping, and giving to NPCs.

Additional mentions, we used Blender for modeling, texturing, unwrapping, rigging and animating. Clip Studio paint for the art and texture design. And finally Audacity for the sound editing.

Challenges we ran into

We ran into the challenges of not having API's available for the vision we had for the game and having to find work arounds with the allotted time. We dealt with a lot of challenges selecting the right AI model fortunately Amazon Bedrock made it easy to change models in our code by easily swapping between gateways. This made it possible for us to try out and sample the different models that AWS had to offer.

Accomplishments that we're proud of

It worked!!! Going into this we had a very ambitious idea for the project. With every set back we would revert to Amazon Q and take a leap forward. Although, we did have to cut several features we are proud of all the features that did make it into the game. Such as a Generative AI that interacts directly with the player, Web extension that ties in directly with the in game Amazon store, and a full proof of concept for this game.

What we learned

Making a game is very difficult. Making a web extension is difficult. Dealing with AI is also very hard. Fortunately using Amazon Q made it a whole lot easier. We learned how to leverage the AWS tools to fast track and progress faster in the Game Development process. We learned how to interact with Amazon Bedrock and incorporate it directly into Unity3D. We learned how to build prompts and pick out the right AI model to fit our needs.

What's next for Delivery Run

Out plan for the future is to fully dive into the Generative AI aspect with the NPC's to bring a richer and more dynamic feel to the game. We would also like to incorporate Generative AI icons and models to make it feel like you are getting the items straight from Amazon. We would like to expand the world and create a variety of different characters with different tasks for the player.