🎬 VerbaVista-AI ( 🌐Live Demo )

YouTube Content Synthesizer using Gemini + LangChain + Streamlit

🚀 Transform any YouTube video into structured notes or an interactive chatbot powered by Google Gemini and LangChain.

---

📖 Overview

VidSynth AI is a Generative AI application that automatically fetches a YouTube video transcript, translates it (if needed), and turns it into:

• 🧠 Structured Notes – AI-generated, topic-wise summaries
• 💬 Interactive Chatbot – Ask questions directly about the video content

It’s built using LangChain, Google Gemini 2.5 Flash, and Streamlit, enabling seamless text extraction, translation, chunking, and contextual question answering — all in one elegant app.

---

✨ Key Features

Feature	Description
🎥 YouTube Transcript Fetching	Automatically extracts video transcripts in multiple languages using YouTubeTranscriptApi.
🌐 Multilingual Translation	Uses Gemini LLM to translate transcripts into English while preserving tone and meaning.
🧩 Chunking & Embeddings	Splits long transcripts and creates embeddings using GoogleGenerativeAIEmbeddings.
💾 Vector Store (RAG)	Stores embeddings in a Chroma vector database for fast, context-based retrieval.
🗂️ AI Notes Generator	Creates structured, human-readable notes using LLM prompting.
💬 Chat with Video	Chatbot mode lets users ask natural language questions about any video content.
🧠 Exponential Backoff Handling	Handles Google API quota limits gracefully with retry logic.

---

🧱 Tech Stack

Category	Tools / Libraries
💡 LLM	Gemini 2.5 Flash Lite via LangChain
🧩 Frameworks	LangChain, Streamlit
🔤 Embeddings	GoogleGenerativeAIEmbeddings
🗄️ Vector DB	Chroma
🎞️ Transcript Extraction	YouTubeTranscriptApi
⚙️ Other Utilities	dotenv, regex, time, google.api_core.exceptions

---

⚙️ Installation & Setup

1️⃣ Clone the Repository

git clone https://github.com/SannidhyaDas/VerbaVista-AI.git
cd VerbaVista-AI

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Add Environment Variables

Create a .env file in the root directory with your Google API Key:

GOOGLE_API_KEY=your_google_api_key_here

🔑 You can get your key from Google AI Studio

🚀 Run the App

streamlit run app.py

Then open your browser at the link Streamlit provides (usually http://localhost:8501).

---

🧩 How It Works — Behind the Scenes

🔹 Step 1: Transcript Extraction - Extracts the video’s transcript (in any supported language) using the YouTubeTranscriptApi.

🔹 Step 2: Translation (Optional) - If the video is not in English, Gemini translates the transcript with cultural and linguistic precision.

🔹 Step 3: Processing Options

• Notes Mode → Extracts key topics and generates structured, concise notes.
• Chat Mode → Creates embeddings, stores them in Chroma DB, and launches a Retrieval-Augmented Generation (RAG) chatbot.

🔹 Step 4: RAG-based Question Answering - When chatting, user queries are matched against the video transcript via embeddings → Gemini answers using only retrieved context.

---

🧰 Key Files

VerbaVista-AI/
│
├── assets/                        # Streamlit web interface
│   ├── appInterface_1.png            # Chat with Video example 
│   ├── appInterface_2.png            # Notes from the video example 
│   └── VerbaVista-pipeline           # working pipeline
│
├── deployment/             # Streamlit deployment setup
│   ├── requirements.txt            # Python dependencies
│   ├── main.py             # Core logic and LLM pipelines  
│   └── app.py              # Streamlit user interface
│
├── localhost/              # setup to run app locally
│   ├── requirements.txt            # Python dependencies
│   ├── main.py             # Core logic and LLM pipelines
│   └── app.py              # Streamlit user interface
│
└── README.md                   # Project documentation

🧠 Example Use Cases

📚 Smart Study Companion - Transform complex academic or lecture videos into clear, structured notes for faster learning and revision.

🎧 Podcast & Interview Analyst - Extract key takeaways and actionable insights from long-form conversations — save hours of manual listening.

🌐 Multilingual Research Assistant - Break language barriers by translating, summarizing, and analyzing global video content in real time.

🏢 Enterprise Knowledge Hub - Turn webinars, product demos, and training sessions into searchable, chat-enabled knowledge bases for internal teams.

💼 Scalable Business Value - Integrate with CRMs or content libraries to automate learning, onboarding, and support, turning video data into searchable, revenue-driving intelligence.

---

🧩 Future Improvements

🎙️ Voice Interaction - Add Speech-to-Text and Text-to-Speech modules for fully voice-based question answering.

🧠 Enhanced Prompt Tuning - Fine-tune Gemini prompts for domain-specific or educational content understanding.

💾 Vector Store Caching - Implement caching for faster reloads and reduced embedding costs.

🧩 Batch Video Summarization - Enable multi-video summarization to process and analyze playlists or course modules efficiently.