TUMany News

Fully automate aggregating and scraping news from RSS, selecting the most relevant ones and produce both full articles and engaging short videos with images and voiceover.

Comment

Inspiration

In today’s world, short-form content—like videos on TikTok, Instagram, and Facebook—dominates how most people consume media. We wanted to not only tackle the challenge but also blend this modern, dynamic way of consuming content with the more traditional experience of reading articles.

What It Does

The program operates fully autonomously and performs the following tasks:

  1. Scrapes and collects articles while cleaning up their content.
  2. Selects and evaluates articles using GPT-4 based on their relevance to electric vehicles (EVs).
  3. Embeds and clusters articles via Ada-02 to identify relevant pieces and related content for enrichment.
  4. Generates enriched articles using multi-step prompt engineering to ensure quality.
  5. Creates automated short videos with voice-over, music, images, and animations.
  6. Delivers content through a specialized platform for easy access and engagement.

Challenges We Faced

  • Scraping content: Many articles have a cluttered DOM structure, making it harder to clean and extract relevant text.
  • Relevance filtering: Identifying articles with genuine relevance turned out to be trickier than expected.
  • Automated video generation: Building a pipeline that produces videos entirely without human intervention proved to be a complex task.

Accomplishments We're Proud Of

We successfully:

  • Collected and processed over 200 articles.
  • Built a fully automated system capable of generating enriched articles by aggregating and summarizing news.
  • Created short videos with voice-overs, animations, music, and other multimedia elements, all without human input.
  • Delivered high-quality, engaging content that we're genuinely proud of.

What We Learned

Through this project, we honed our skills in:

  • Leveraging OpenAI APIs for scraping, article generation, and prompt-tuning.
  • Cleaning and processing articles efficiently.
  • Automating video generation.
  • Designing and mocking up user-friendly front-end interfaces.

What's Next for TUMany News

Looking ahead, we aim to:

  1. Improve the project’s stability and reliability.
  2. Enhance the video generation process, which was both challenging and enjoyable to develop.
  3. Expand the platform to include additional features and integrate it into a broader ecosystem of tools and resources.
Share this project:

Updates