The Narrator

Inspiration Many of our favorite childhood stories are passed down through generations with word-of-mouth or oral traditions - As a means of empowering and preserving the tales and those who narrate them, we created a platform for individuals to record their favorite stories to transform them into storybooks with a visual representation. We placed a particular emphasis on sharing stories with everyone, and ensured the implementation of a read-out-loud feature to narrate the tale for those who may be unable to read or be visually impaired. Our goal was to promote inclusivity and empower the preservation of generational storytelling.

Product

Our website, the Narrator, takes in a story through narration or text, to generate a shareable and virtual picture book with live narration

Creation Process

Frontend: We utilized Nextjs, for the simplicity of routing and deployment. Nextjs provides a framework that makes React more effective to work with.

Database

We used MongoDB to store user collections and their created stories. We communicated with MongoDB through the package Mongoose. This provided an efficient SDK for interacting with the database. We chose to define a route in the Nextjs as an API route, which served to write to our database for the necessary information.

API’s

We mainly focused on the GPT-3.5 model. This allowed us to process paragraphs to break them down into smaller chunks for Dalle AI to generate images. We used the Dalle AI to generate images, speak to text (vice versa).

Challenges

The initial set-up posed a challenge with the adaption of new technologies. Communication with MongoDB using Nextjs was difficult - Rather than using the SDK Mongoose, we had to work though integration (defining API routes, calling them). In addition, we struggled to initiate the Chat-3.5 model and had to fine tune some of our existing parameters to split the processed text. Image storing with the lack of a storage bucket was a challenge as well, but the utilization of the Dalle image URL rather than chat was a solution. Should we have had more time, we would have preferred to implement a storage bucket.

Accomplishments

Through this process, we are most proud of the deployment of our application to the public. Knowing that others can enjoy our app and preserve their traditions in the real-world is truly a rewarding experience. We are excited to see this happen in real time, seeing that this was the first hackathon where this was successful.

Lessons

The main lessons we learnt through our work in this hackathon regarded how image generation and speech detection work and how standardized data is generated. In addition, we learnt the process of deploying a web application in its initial stages.

Moving Forward

Along with our goal of generational story preservation, we hope to implement real-time translation to all parts of our product in the future. Accessibility is a huge feature that we placed an emphasis on during our creation stage, and we are currently working on taking in more than one language as an input (languages other than English). Being aware that generation storytelling is a global tradition, we plan to provide inclusive language options in our next version.

Built With

• bootstrap
• chatgpt-api
• dalle
• github
• javascript
• mongodb
• mongoose
• next.js
• openai
• react
• speakly-api
• typescript
• vercel