EchoFact

EchoFact is a real-time journalist tool that provides immediate insights into the veracity of claims, furnishes contextual understanding, and suggests avenues for further investigation dynamically

Comment

Inspiration

In a world where misinformation, half-truths, and outright lies can spread instantly—even from the highest levels of government—journalists, researchers, and the public face an unprecedented challenge: how do you counter falsehoods as they happen? The stakes are enormous, from public trust in institutions to the integrity of democratic processes.

EchoFact was inspired by the urgent need for a real-time truth defense system—a tool that equips journalists and fact-finders to challenge and verify claims the moment they’re made. By combining live speech transcription and AI-powered fact-checking, EchoFact enables its users to expose inaccuracies, provide context, and surface credible sources with speed and confidence, no matter the information battlefield.

What it does

EchoFact is a web application that combines live speech transcription with AI-driven fact-checking to arm users against misinformation in real time. Whether monitoring a live press conference, analyzing a news broadcast, or reviewing political debates, EchoFact captures every spoken or written claim, analyzes it with Perplexity’s Sonar AI, and delivers a clear verdict: True, False, Misleading, or Needs Context. Each claim is backed by transparent explanations and direct links to credible sources, making it easy to share and cite the truth instantly.

How we built it

  • Frontend: React + TypeScript single-page app for a fast, seamless, and responsive experience.
  • Speech Recognition: Leverages the Web Speech API for real-time audio transcription, instantly segmenting claims as they’re spoken.
  • AI Fact-Checking: Integrates the Perplexity Sonar API to verify each claim, allowing users to add custom context for more nuanced analysis.
  • User Experience: Designed for accessibility, the app uses Tailwind CSS and custom theming for light, dark, and high-contrast modes. Fact-checks and transcripts are clearly separated, with intuitive status icons and filtering.
  • Session & Security: User API keys are managed locally for privacy, and sessions can be reset in one click.

Challenges we ran into

  • Accurate claim detection in spontaneous speech—identifying which statements should be fact-checked without overloading users.
  • Structuring AI prompts for speed, relevance, and clarity in verdicts and explanations.
  • Designing an interface that balances rapid delivery of insights with clear, accessible presentation for all users.
  • Handling errors gracefully across both transcription and AI services, to maintain user trust in high-pressure scenarios.

What we learned

  • Misinformation is rarely black and white; providing context and source credibility is just as important as the verdict itself.
  • Real-time tools are only as useful as their speed and clarity—instant feedback empowers journalists to act, while detailed explanations and source transparency ensure their confidence.
  • Customizability and accessibility aren’t nice-to-haves—they’re mission-critical for journalists working in diverse, high-stress environments.

What’s next

  • Improving claim detection with advanced NLP, so only the most important, checkable statements are flagged.
  • Building collaboration features so newsroom teams can work together on live fact-checking.
  • Developing specialized modules for legal and journalistic workflows, with tailored information retrieval and reporting styles.
  • Integrating additional knowledge sources (legal, scientific, etc.) for even broader verification power.

Built With

Share this project:

Updates