AI-Sanj: Unmasking Digital Deception with Gemini 3 The Inspiration This app addresses an increasingly serious social issue that has emerged with the rapid growth of artificial intelligence. Today, AI-generated and AI-edited images, videos, and voice content can appear highly realistic, making it difficult for ordinary users to distinguish between real and manipulated media. This often leads to misunderstandings, damaged friendships, and, in some cases, serious family conflicts. We built AI-Sanj to give people a tool to verify the truth and protect their digital dignity.
What it does AI-Sanj is a multimodal forensic prototype designed to provide a cautious and initial assessment of suspicious digital content. It doesn't just look for AI; it looks for human manipulation too.
Image Analysis: Detects "cut-and-paste" edits, mismatched lighting, and unnatural skin textures.
Video Forensics: Monitors biological signs like inconsistent blinking and lip-sync errors.
Audio Analysis: Identifies frequency shifts in cloned voices and the absence of natural emotional breathing patterns.
The Role of Gemini 3 Gemini 3 serves as the core analytical engine. Utilizing its advanced multimodal reasoning, the app identifies forensic patterns commonly associated with AI-generated or manually edited media. Instead of a simple "Real or Fake" answer, Gemini 3 provides a structured, step-by-step explanation, helping users understand why a piece of content is suspicious. Beyond analysis, Gemini 3 also played a key role as a development assistant—supporting script creation, improving logical flow, and refining the overall design for a seamless user experience.
Technical Implementation The app is intentionally maintained as a testable setup within Google AI Studio and Streamlit. It relies on a user-provided Gemini API key to demonstrate its full functionality in real-time. By combining AI power with human-centric design, we've created a solution that is both powerful and easy to understand for the average user.
Our Vision This project promotes digital awareness, caution, and the responsible use of AI. It demonstrates how advanced models like Gemini 3 can be applied to protect trust and social relationships rather than undermine them. Our goal is not just to deliver a verdict, but to help users pause, reflect, and make informed decisions in a world of digital doubt.
Built With
- gemini-3-api
- google-ai-studio
- pil
- python
- streamlit
Log in or sign up for Devpost to join the conversation.