Further Information

Inspiration: I knew that literacy rates in the United States have dropped below 80% last year, and that only 54% of adults can read at a 6th grade level. Also, I have received the advice "Visualize what you are reading" too often, and so I stumbled on the idea of generating images automatically for people in an attempt to boost reading comprehension.

What it does: Word Canvas is a website that uses state-of-the-art machine learning models to provide images for a text real-time as an aid.

How I built it: The website's user interface is coded in index.html and story.html, styled with app.css. The machine learning part in ml.py uses ChatGPT 3.5 and Dall-E 2 APIs to summarize text and generate images, seamlessly integrated into the website through app.py.

Challenges I ran into: The main error that I kept getting had to do with the prompts that I was entering into the APIs. It is a known problem that Dall-E 2, or image generation models in general, have a problem including text into their images as it often turn illegible. However, my pictures were often containing text, which is an issue that took a lot of troubleshooting to get around.

Accomplishments that I'm proud of: I am proud of the user interface because to me, it looks clean and very professional. Additionally, I am proud of the accuracy of the images in relation to the text.

What I learned: This is my first time using the ChatGPT and Dall-E 2 API, so I learned a lot about how it is implement and how to prompt these APIs to get exactly what I needed.

What's next for Word Canvas: In the future, I plan to make a database where people can log in and have an account for this website. They should be able to download their own text as a pdf, and the website should be able to read it and set up a page for it.

Built With

• chatgpt
• css
• dall-e2
• flask
• html
• javascript
• python