This is a submission for the Cloudflare AI Challenge.
What I Built
The app I built is designed to create personalized audio podcasts based on the latest news and events tailored to the user’s specific interests.
By selecting their preferred topics, the app curates news articles and other relevant content from various sources and converts them into an audio podcasts.
Users can then listen to these custom podcasts while multitasking, such as during their workout routine, traveling, or performing household chores. The goal is to provide a convenient and engaging way for users to stay informed about the topics they care about most.
Demo
My Code
Tech Stack :
Next js
Cloudflare Pages
Workers AI API
Journey
The motivation behind this app stemmed from the realization that staying informed about the latest news and events can be a challenge in our fast-paced, multitasking world. With an abundance of information available, it’s easy to feel overwhelmed or miss out on important updates, especially if they’re not directly related to one’s specific interests.
This app tackles several problems:
Information Overload: By curating and summarizing news articles based on user interests, the app helps users avoid information overload and focus on the topics that truly matter to them.
Accessibility: Converting news content into an audio format allows users to consume information while multitasking, such as during their commute, workout, or household chores, making it easier to stay informed without sacrificing productivity.
Looking ahead, I envision expanding the app’s capabilities to include interactive features, such as the ability to provide feedback or engage with the content through voice commands. Additionally, I plan to explore partnerships with major news organizations and content providers to further enhance the depth and breadth of the app’s content offerings.
I’m really feeling accomplished after building NewsCast in this short span of time. I got to know about the Cloudflare AI challenge a bit late (i.e., 6th April), and I started to code on 9th April. Along with this, I was completely new to Cloudflare and serverless architecture. So, because of the challenge, I learned alot of new things and solved a problem that has been on my mind for a long time.
Multiple Models and/or Triple Task Types
Models –
@cf/openai/whisper
@cf/facebook/bart-large-cnn
Task Types –
Summarization: To provide concise and informative podcast content, news articles are processed through a summarization model.
Speech Recognition: Once the podcast script is generated, a speech recognition model is used to create accurate captions for the audio version, enhancing accessibility and allowing users to easily follow along or reference specific sections.
Translation [Pending to add]: To cater to a global audience, the generated podcast scripts can be translated into various languages, enabling users to consume the content in their preferred language.
Text Generation: At the core of the app is a text generation model that takes the summarized news articles and user interests as input, and generates a cohesive and engaging script for the audio podcast.
