VidPhrase

Search phrases inside YouTube videos, uploaded video and audio files using subtitles, comments, descriptions, and AI-powered semantic search.

✨ Features • 🧠 How It Works • 🛠️ Tech Stack • ⚙️ Setup • ⚠️ Cookies Helper • ▶️ Usage • 📁 Structure

---

✨ Features

• 🔎 Search phrases inside YouTube videos
• 🎬 Search inside uploaded video files
• 🎧 Search inside uploaded audio files
• 📝 Search using YouTube subtitles
• 💬 Search inside YouTube comments
• 📄 Search inside video descriptions
• 🧠 AI-powered semantic search
• 🌍 Multi-language subtitle support
• 🎙️ Local transcription using Faster-Whisper
• 📥 Download subtitles as text files
• ⚡ Fast and lightweight Flask web interface

---

🧠 How It Works

VidPhrase supports multiple search methods.

🔍 Exact & Fuzzy Search

Search directly inside:

• YouTube subtitles
• YouTube comments
• Video descriptions
• Uploaded video transcriptions
• Uploaded audio transcriptions

Even if the phrase contains typos or slightly different wording, fuzzy matching can still return relevant results.

🤖 AI Semantic Search

VidPhrase can search by meaning, not only by exact keywords.

The AI understands:

• synonyms and paraphrases
• abbreviations and acronyms
• broader and narrower concepts
• technologies, products, services, and brands related to the query
• explanations and examples that imply the same idea
• surrounding context even when the exact words never appear

Example

Query

artificial intelligence

Possible matches:

• AI
• Machine Learning
• Deep Learning
• Neural Networks
• LLMs
• ChatGPT
• GPT
• Gemini
• Qwen

---

Query

cloud technologies

Possible matches:

• AWS
• Amazon Web Services
• Google Cloud
• Azure
• Kubernetes
• Docker
• Containers
• Serverless
• Cloud Infrastructure
• Virtual Machines

This allows you to discover moments that are conceptually related to your query instead of being limited to exact keyword matching.

---

🛠️ Tech Stack

• 🐍 Python
• 🌐 Flask
• 📺 yt-dlp
• 🎙️ faster-whisper
• 🔍 thefuzz
• 🤖 Google GenAI
• 🍪 browser-cookie3
• 🔧 Werkzeug

---

⚙️ Setup

1️⃣ Install Python

Windows

Download Python from:

https://www.python.org/downloads/

During installation, make sure to enable:

• ✅ Add Python to PATH
• ✅ Install pip

Ubuntu / Debian

sudo apt update
sudo apt install python3 python3-pip

---

2️⃣ Verify Installation

Windows:

python --version
pip --version

Linux:

python3 --version
pip3 --version

---

3️⃣ Clone the Repository

git clone https://github.com/yourusername/VidPhrase.git
cd VidPhrase

---

4️⃣ Create a Virtual Environment (Recommended)

python -m venv venv

Windows

venv\Scripts\activate

Linux / macOS

source venv/bin/activate

---

5️⃣ Install Dependencies

pip install -r requirements.txt

or

pip3 install -r requirements.txt

---

6️⃣ Configure Gemini API

VidPhrase uses Google Gemini for semantic search.

Open app.py and replace:

client = genai.Client(api_key="GEMINI_API_TOKEN")

with your own API key:

api_key="YOUR_GEMINI_API_KEY"

Without a valid Gemini API key, AI Semantic Search will not work.

---

⚠️ Cookies Helper

YouTube sometimes rate-limits requests and yt-dlp may fail with:

HTTP Error 429: Too Many Requests

If this happens, generate fresh browser cookies by running:

python3 cookie_fetch_profiles.py

The script automatically extracts YouTube cookies from available Chrome and Firefox profiles and saves them inside the cookies/ directory.

After generating the cookies, restart VidPhrase and try again.

---

7️⃣ Run the Application

Windows:

python app.py

Linux:

python3 app.py

Open:

http://127.0.0.1:9005

---

▶️ Usage

1. Open VidPhrase in your browser.
2. Paste a YouTube link or upload a video/audio file.
3. Enter the phrase you want to find.
4. Choose the search method:

• 📝 Subtitle Search
• 💬 Comment & Description Search
• 🎙️ Whisper Transcription Search
• 🧠 AI Semantic Search

5. Browse the results.
6. Jump directly to the relevant moment and continue from the surrounding context.

---

📁 Project Structure

VidPhrase/
├── app.py
├── cookie_fetch_profiles.py
├── requirements.txt
├── cookies/
│   ├── cookies_1.txt
│   ├── cookies_2.txt
│   ├── cookies_3.txt
│   ├── cookies_4.txt
│   └── cookies_fire.txt
├── templates/
├── static/
└── README.md

---

📝 Notes

• re
• os
• json
• glob
• tempfile
• uuid
• shutil
• warnings
• io.BytesIO

These modules are part of Python's standard library and do not need to be installed separately.

• Subtitle search uses fuzzy matching and may return closely related phrases.
• AI search is designed to find contextual and conceptual matches, not only exact words.
• Local video and audio files are automatically transcribed using Faster-Whisper.
• The application runs on port 9005 by default.

---

Made with ❤️ to make searching through videos and audio effortless.