🔊 Chat with Your YouTube Video using LLMs 🤖

🚀 Built an AI-powered app that lets you chat with YouTube videos using LLMs!
I recently developed ChatAudio, a project that allows users to interact with YouTube video content in a conversational way — powered by LangChain, OpenAI, and Streamlit.
🎯 Goal
Enable users to extract transcripts from YouTube videos and ask any question about the content — just like chatting with the video!

🔁 Workflow Overview
1️⃣ User inputs a YouTube video URL
2️⃣ App fetches subtitles using YouTubeTranscriptAPI
➕ If subtitles are missing, you can optionally use AssemblyAI for speech-to-text
3️⃣ Transcript is saved and processed via LangChain:
Loaded with TextLoader
Split using CharacterTextSplitter
Embedded via OpenAIEmbeddings
Stored with FAISS for fast vector search
4️⃣ The query is passed to the LLM (ChatOpenAI) using a RetrievalQA chain
5️⃣ 💬 User receives a smart answer based on actual video content!

🧠 Tech Stack
✅ Streamlit – UI & interaction
✅ LangChain – Document processing & RAG
✅ OpenAI GPT-3.5 – Question answering
✅ FAISS – Vector storage
✅ YouTubeTranscriptAPI – Subtitle extraction
✅ AssemblyAI (Optional) – Audio transcription fallback
✅ PyTube – For downloading audio from YouTube

📌 Key Highlights
✔️ Converts YouTube content into a conversational AI interface
✔️ Real-time interaction with a smooth Streamlit frontend
✔️ Skip the full video – just ask what you want to know
✔️ Modular codebase with robust error handling and retrieval pipeline
✔️ Efficient, fast, and scalable for real-world use

🐳 Docker Deployment
To make it easy for anyone to run the app locally, I’ve also Dockerized the project and published it on
Docker Hub: https://lnkd.in/d-trNx98