About WackyTalk AI
Inspiration
The inspiration for WackyTalk AI came from the "Useless Sh!t" challenge and a desire to create something purely entertaining and unexpected. I wanted to combine the vast, often humorous, content of Reddit with the emerging capabilities of AI video generation to create a truly unique and shareable experience. The idea was to build an app that serves no practical purpose but brings a smile to your face.
What I Learned:
This project has been a fun exploration of creative API integrations and frontend development for entertainment. Key learnings include:
Reddit API Interaction: Understanding how to fetch public data from the Reddit API, including handling user agents and parsing JSON responses to extract relevant content (like post titles or self-text).
Tavus AI Video Generation: Deepening my understanding of how to use Tavus's API to generate dynamic video content from text, including selecting actors and managing the asynchronous nature of video creation.
Frontend Development for Engagement: Focusing on UI/UX elements that enhance the "fun" factor, such as clear buttons, loading indicators, and a straightforward user flow for generating and playing content.
Web Sharing Mechanisms: Implementing simple browser-based sharing functionalities to platforms like Reddit, enabling users to easily spread their unique experiences.
How I Built My Project:
WackyTalk AI is a React web application, leveraging Tailwind CSS for its quirky and engaging design. The InterviewGenerator component is the heart of the application.
Project Setup: Initiated a standard React web app via Bolt.new and set up Tailwind CSS.
Reddit Question Fetching: Implemented a fetch call to the Reddit API to pull random, absurd questions from subreddits like r/showerthoughts.
Tavus AI Video Integration: Developed the logic to send the fetched question to the Tavus API for AI video generation, then embedded the resulting video URL for playback.
Share Functionality: Added a "Share to Reddit" button that constructs a pre-filled Reddit post URL, allowing users to easily share their interview experience.
The focus was on creating a seamless and entertaining flow from question generation to AI video playback and sharing.
Challenges I Faced:
Reddit API Nuances: Navigating Reddit's API rules and understanding how to fetch public data efficiently without hitting rate limits or requiring complex OAuth for a simple read-only use case.
Tavus Video Latency: AI video generation can take some time. Managing the loading states and providing clear user feedback during this process was important for a good user experience.
Content Filtering: Ensuring that the fetched Reddit content was appropriate and truly "absurd" (and not offensive) required some basic filtering or careful subreddit selection.
Deployment & Public Sharing: While the sharing button is implemented, the actual ability for others to view the generated videos would depend on having a publicly accessible hosting solution for the app itself. I did not implement this yet
Accomplishments that we're proud of
- Wacky Talk was started with a goal of developing three different apps during the competition while also working full time and having 5 kids. I am proud of what I was able to create with Tavus in the sort timeframe and other constraints even if its not perfect yet.
What's next for Wacky Talk
- Added the hosting for videos so they can be shared
- Adding ability for users to converse with the AI agent about the silly questions
Log in or sign up for Devpost to join the conversation.