DeepFake Detector AI

• 

  DeepFake Detector AI - Homepage

Inspiration

As AI-generated content becomes more realistic, it's harder for everyday users to know what’s real and what’s fake. Scammers are using deepfake voices and videos to trick people, especially in vulnerable communities. We wanted to build a solution that detects this in real time—while also teaching people how to protect themselves.

What it does

DeepFake Detector AI is an AI-powered platform that lets users:

• Upload videos, images, or audio and detect if they’re AI-generated or real.
• Have real-time voice conversations with an AI expert.
• Watch short, AI-generated video lessons on detecting scams and deepfakes.
• Ask their own questions and receive multilingual AI responses (English, Hausa, Arabic, French).

How we built it

We used:

• Bolt.new to build the full stack, UI/UX, and workflows.
• Tavus to generate AI video responses and expert lessons.
• ElevenLabs for real-time voice conversations with expert agents.
• Netlify for deployment and hosting.
• Multi-language support for regional inclusiveness.
• Custom prompts to power the “Ask an Expert” voice and video interface.

Challenges we ran into

• Real-time voice conversations required deep troubleshooting to handle mic permissions, WebSocket latency, and feedback loops.
• Creating natural, human-like Tavus video agents that felt trustworthy.
• Designing UI/UX that is both educational and engaging across all languages and devices.

Accomplishments that we're proud of

• Successfully integrated Tavus for conversational AI video responses that feel human and engaging.
• Enabled real-time voice conversations using ElevenLabs, allowing users to speak directly with an AI expert agent in multiple languages.
• Built a clean, accessible, and mobile-friendly user interface that adapts well for global users, including support for Hausa, Arabic, French, and English.
• Launched a deepfake detection engine for video, audio, and images — all within a single platform.
• Created an “Ask an Expert” tab where users can either click pre-set questions or ask their own, and get real-time video + voice responses.
• Completed full deployment via Netlify with a free custom domain from Entri.
• Designed a fully functioning MVP within the hackathon timeframe using Bolt.new — from idea to product.

What we learned

We learned how to combine cutting-edge AI tools to build something powerful and inclusive. This project taught us how to balance tech with trust, and we’re proud to launch it as a tool to make the world safer from AI scams.

What's next for DeepFake Detector AI

• Expand to mobile app: We plan to build a native app version for Android and iOS.
• Community awareness campaign: Launch multilingual campaigns targeting youth, journalists, and vulnerable communities.
• Partner with fact-checkers and educators to include our platform in their digital literacy curriculum.
• Improve detection accuracy by integrating with more advanced deepfake detection models and fine-tuning datasets.
• Enable user submissions of suspicious content to help us build an open database of real vs. AI-generated material.
• Launch "Expert Personas" so users can choose different AI experts (e.g., cybersecurity analyst, digital rights activist, scam survivor) to respond in varied tones.
• Introduce API Access so others can embed DeepFake Detector AI into their apps, browsers, and educational platforms.

Built With

• arabic
• bolt.new
• elevenlabs
• hausa
• javascript
• multilingual-support
• netlify
• react
• tavus