Inspiration

We realized that we were often tired of watching the same types of media over and over again. Because it can be difficult to explore your boundaries, we decided to create a tool meant to help people explore different types of media.

What it does

Our project recommends certain media to a user after they answer some questions like what genre or themes they would like to see. After some time, an AI generates some media that the user can look up and watch. Some of the questions will vary depending on the medium. (For example, video games will ask for difficulty while movies will ask for actors.)

How we built it

We first took some time to think about what mediums we think about. We settled on 5: video games, movies, tv shows, books, and music. Afterwards, we then asked ourselves what questions a user may ask, such as if they want works similar to certain artists, a certain length, if they thought plot was important, and much more. We settled on using OpenAI to generate our sources because its API seems to be the best to implement for our situation. We then integrated the service into our website, using the API to generate responses after we gave additional parameters based on the media type and what else the user would want. Finally, we added some style to our website, polishing the fine details.

Challenges we ran into

OpenAI has some limitations, such as not generating responses if too many calls to the API are made in a short time. As such, we tried to implement a delay between requests so the user would not encounter these problems.

Accomplishments that we're proud of

We're proud of learning how to integrate an AI service into a project and use it to the best of our ability.

What's next for Retro Recommend

We would like to see if we can improve the quality of ChatGPT's responses, including:

  • Seeing if we can add more parameters or questions
  • Getting around the limit of calls within a certain time frame
  • If different wordings of the same questions would result in more accurate answers.

We also wish to improve some fine details of our project, like cleaning up the UI further.

Built With

Share this project:

Updates