Dilkush
Dilkush: AI-Powered Language Learning Workspace
A Vue.js web application that transforms YouTube videos into interactive language learning experiences using Google’s Gemini API.
🎯 Project Overview
Dilkush is a language learning platform designed to streamline Turkish exam preparation by automatically generating listening comprehension questions from YouTube videos. What started as a manual process of downloading videos, transcribing them through third-party services, and manually creating questions with ChatGPT, evolved into an automated, intelligent workspace.
Live Demo: https://dilkush.web.app
🚀 Key Features
- YouTube Video Processing: Direct URL input for automatic question generation
- Interactive Learning: Answer questions and receive instant feedback
- Adaptive Reading Materials: Search and adjust content based on proficiency level (e.g., B2)
- Exam Preparation Focus: Designed specifically for Turkish language certification tests
🛠️ Technical Stack
- Frontend: Vue.js
- Backend: Firebase
- AI Integration: Google Gemini API for video understanding
- Framework: Firebase Genkit for serverless functions
💡 The Problem & Solution
Challenge: Traditional language exam preparation required a multi-step manual process:
- Download YouTube videos
- Upload to transcription services (Speechmatics, ElevenLabs)
- Copy transcripts to ChatGPT with custom prompts
- Manually format questions
Solution: Leveraged Google Gemini’s native YouTube video understanding capabilities to automate the entire workflow through a single URL input.
🔮 Future Roadmap
- Writing Module: AI-powered writing exercises and feedback
- Speaking Assessment: Voice recognition and pronunciation evaluation
- Cultural Content: Turkish idioms, proverbs, and cultural references
- Comprehensive Exam Simulation: Full Turkish proficiency test experience
🎓 Learning Outcomes
This project allowed me to explore:
- Google Gemini API’s multimodal capabilities
- Firebase Genkit framework (first-time implementation)
- Vue.js frontend development
- AI-driven educational tool design
*Note: This is an early-stage project. Please use gently as testing is