RT-Vision-Core: Video Transcription for Obsidian

A powerful tool to transcribe YouTube videos and Instagram posts/reels into Obsidian-compatible markdown notes using OpenAI's Whisper AI.

Features

• YouTube Support: Download and transcribe any YouTube video
• Instagram Support: Download and transcribe Instagram posts and reels
• AI Transcription: Uses OpenAI Whisper for high-quality transcription
• Multiple Languages: Auto-detects language or specify manually
• Obsidian Integration: Creates formatted markdown notes ready for your vault
• Timestamps: Optional timestamp support for detailed notes
• Batch Processing: Process multiple URLs in one command
• Flexible Models: Choose from 5 Whisper models based on speed vs accuracy needs

Installation

Prerequisites

• Python 3.8 or higher
• ffmpeg (for audio/video processing)

Install ffmpeg

macOS:

brew install ffmpeg

Ubuntu/Debian:

sudo apt update
sudo apt install ffmpeg

Windows: Download from ffmpeg.org or use:

winget install ffmpeg

Setup

1. Clone the repository:

git clone <repository-url>
cd RT-vision-core

2. Create a virtual environment (recommended):

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Configure (optional):

cp .env.example .env
# Edit .env to configure Whisper model, language, and Obsidian vault path

Usage

Basic Usage

Transcribe a YouTube video:

python main.py https://www.youtube.com/watch?v=dQw4w9WgXcQ

Transcribe an Instagram post/reel:

python main.py https://www.instagram.com/p/ABC123/

Process multiple URLs:

python main.py https://www.youtube.com/watch?v=video1 https://www.instagram.com/reel/xyz/

Advanced Options

Include timestamps in transcription:

python main.py --timestamps https://www.youtube.com/watch?v=dQw4w9WgXcQ

Use a different Whisper model:

python main.py --model medium https://www.youtube.com/watch?v=dQw4w9WgXcQ

Specify language (skip auto-detection):

python main.py --language es https://www.youtube.com/watch?v=dQw4w9WgXcQ

Command Line Options

• urls: One or more YouTube or Instagram URLs (required)
• --timestamps: Include timestamps in the transcription
• --model {tiny,base,small,medium,large}: Whisper model to use (default: base)
• --language LANG: Language code for transcription (e.g., en, es, fr)

Whisper Models

Choose based on your needs:

Model	Speed	Accuracy	Memory	Best For
tiny	⚡⚡⚡⚡⚡	⭐⭐	~1 GB	Quick drafts
base	⚡⚡⚡⚡	⭐⭐⭐	~1 GB	Balanced (default)
small	⚡⚡⚡	⭐⭐⭐⭐	~2 GB	Good quality
medium	⚡⚡	⭐⭐⭐⭐⭐	~5 GB	High accuracy
large	⚡	⭐⭐⭐⭐⭐	~10 GB	Best quality

Output

Transcriptions are saved as markdown files in the transcripts/ directory with the following format:

---
title: Video Title
source: https://www.youtube.com/watch?v=...
date: 2025-11-10
language: en
channel: Channel Name
platform: YouTube
tags:
  - transcription
  - en
---

# Video Title

## Metadata

- **Source:** https://www.youtube.com/watch?v=...
- **Channel:** Channel Name
- **Date:** 2025-11-10
- **Duration:** 15 minutes
- **Language:** en

## Description

Original video description...

## Transcription

Transcribed text goes here...

Obsidian Integration

If you set OBSIDIAN_VAULT_PATH in your .env file, notes will be automatically copied to your Obsidian vault:

OBSIDIAN_VAULT_PATH=/path/to/your/obsidian/vault

Project Structure

RT-vision-core/
├── main.py                    # CLI entry point
├── config.py                  # Configuration settings
├── youtube_downloader.py      # YouTube video/audio download
├── instagram_downloader.py    # Instagram post/reel download
├── transcriber.py             # Whisper transcription
├── obsidian_formatter.py      # Markdown formatting
├── requirements.txt           # Python dependencies
├── .env.example              # Example environment config
├── downloads/                # Temporary download directory
└── transcripts/              # Output markdown files

Examples

YouTube Video with Timestamps

python main.py --timestamps --model small https://www.youtube.com/watch?v=dQw4w9WgXcQ

This will create a note with timestamped segments:

## Transcription

**[00:00]** Welcome to this video about...
**[00:15]** In this tutorial, we'll cover...
**[00:45]** First, let's talk about...

Instagram Reel in Spanish

python main.py --language es https://www.instagram.com/reel/ABC123/

Batch Processing

python main.py \
  https://www.youtube.com/watch?v=video1 \
  https://www.youtube.com/watch?v=video2 \
  https://www.instagram.com/p/post1/ \
  https://www.instagram.com/reel/reel1/

Troubleshooting

Common Issues

1. ffmpeg not found: Install ffmpeg using instructions above
2. Out of memory: Use a smaller Whisper model (tiny or base)
3. Instagram download fails: Some private or restricted posts may not be accessible
4. Slow transcription: Use a smaller model or ensure you have GPU support

GPU Acceleration

For faster transcription, install PyTorch with CUDA support:

# For NVIDIA GPUs
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is open source and available under the MIT License.

Acknowledgments

• OpenAI Whisper for transcription
• yt-dlp for YouTube downloads
• Instaloader for Instagram downloads

Support

For issues and questions, please open an issue on GitHub.