VibeCheck

• 

  VibeCheck

Inspiration

One wrong upload can turn fame into fallout. I’ve seen creators lose their audience overnight because of a misunderstood tone or risky joke. I built VibeCheck to prevent that. It gives YouTube creators a smart pre-publish reality check before the internet does - scanning tone, clarity, and controversy using AI.

What It Does

VibeCheck is a Chrome extension that analyzes YouTube videos before they go live. It reviews titles, descriptions, transcripts, and thumbnails to detect sentiment, clarity, reputation risk, and clickbait or misleading imagery.

Creators can paste text, upload transcripts, drop a YouTube link, or simply speak their description aloud. Speech is instantly transcribed to text using the Web Speech API and analyzed in real time, making content review faster and easier than ever.

How I Built It

I built VibeCheck as a Chrome extension using Manifest V3, designed with a dual-mode interface: a compact popup for quick checks and a full-page app for detailed analysis.

Frontend: Vanilla JavaScript (ES6+), HTML5, CSS3 On-device AI: Gemini Nano via Chrome’s Prompt API Cloud fallback: Google Gemini 2.0 Flash API for multimodal analysis Additional APIs: Web Speech API for speech-to-text, YouTube oEmbed API for metadata and transcript extraction

The background service worker routes messages between the UI and the hybrid AI engine. The system automatically detects whether Chrome’s built-in AI is available and switches to the cloud model when needed.

Challenges

Integrating Chrome’s on-device Gemini Nano with the Gemini Flash API in a seamless hybrid setup was a major challenge. Building real-time voice transcription and maintaining consistent UX across popup and full-page modes required careful state management. Ensuring privacy, performance, and reliability under the constraints of Manifest V3 was also a key engineering focus.

What I Learned

I learned how to combine on-device and cloud AI into a single intelligent workflow. Working with Chrome’s new built-in AI APIs gave me a deeper understanding of privacy-preserving machine learning and edge inference. I also refined my approach to UI responsiveness, asynchronous messaging, and multimodal analysis in browser extensions.

What’s Next

Next, I plan to expand VibeCheck with emotion-level scoring (sarcasm, aggression, positivity), support for short-form content such as YouTube Shorts, and team-level brand safety reporting. I also aim to publish it on the Chrome Web Store as a free AI safety and content quality tool for creators.

Built With

• api
• chrome
• chrome-extensions-(manifest-v3)
• css3
• gemini-nano-(chrome-built-in-ai)
• google-gemini-2.0-flash-api
• html5
• javascript-(es6+)
• web-speech-api-(voice-input-and-transcription)
• youtube-oembed-api-(metadata-and-transcript-extraction)