https://github.com/nfs-tech-bd/subgenai
SubgenAI is a command-line tool that uses OpenAI's Whisper model to transcribe video audio and generate SRT subtitles. It supports multiple models for varying speed/accuracy, optional translation to English, and provides progress tracking with timestamps. Simple to use and flexible for video-to-subtitle conversion.
https://github.com/nfs-tech-bd/subgenai
ai automation cli ffmpeg machinelearning openai openai-whisper speechrecognition srt-subtitles subtitlegenerator subtitles transcription translation videoprocessing whisper whisper-ai
Last synced: 4 days ago
JSON representation
SubgenAI is a command-line tool that uses OpenAI's Whisper model to transcribe video audio and generate SRT subtitles. It supports multiple models for varying speed/accuracy, optional translation to English, and provides progress tracking with timestamps. Simple to use and flexible for video-to-subtitle conversion.
- Host: GitHub
- URL: https://github.com/nfs-tech-bd/subgenai
- Owner: nfs-tech-bd
- Created: 2025-10-01T09:06:13.000Z (5 days ago)
- Default Branch: main
- Last Pushed: 2025-10-01T10:01:56.000Z (5 days ago)
- Last Synced: 2025-10-01T11:27:53.184Z (5 days ago)
- Topics: ai, automation, cli, ffmpeg, machinelearning, openai, openai-whisper, speechrecognition, srt-subtitles, subtitlegenerator, subtitles, transcription, translation, videoprocessing, whisper, whisper-ai
- Language: Python
- Homepage:
- Size: 138 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 🎬 SUBGEN AI — Subtitle Generator (CLI)
---
## 🚀 Overview
**SubgenAI** is a lightweight and efficient **command-line tool** that uses **OpenAI's Whisper model** to automatically transcribe and generate subtitles (`.srt`) from video files.
✨ With a few simple commands, you can:
- Extract audio from any video
- Generate accurate subtitles
- Optionally translate them into **English**
- Save directly as `.srt` with automatic or custom file naming
---
## ✨ Features
- 🎯 **Simple CLI** – Run directly from your terminal
- 📝 **Automatic Naming** – Saves as `video_name.srt` if no output is specified
- 🌍 **Translate to English** – From any source language
- ⚡ **Clean & Efficient** – Uses temporary storage and cleans up automatically
- 🔄 **Flexible Models** – Choose from `tiny`, `base`, `small`, `medium`, `large`
---
## 🛠️ Tech Stack
- **Language**: Python
- **Transcription**: OpenAI Whisper
- **Audio Extraction**: FFmpeg
---
## 📦 Prerequisites
Before using SubgenAI, make sure you have:
- 🐍 **Python 3.8+**
- 🎵 **FFmpeg** (required for audio extraction)
### 🔧 Installation Instructions
- **Windows**: Download from the official [FFmpeg site](https://ffmpeg.org/download.html) and add the `bin` folder to PATH
- **macOS** (Homebrew):
```bash
brew install ffmpeg
```
- **Linux** (apt):
```bash
sudo apt update && sudo apt install ffmpeg
```
---
## ⚙️ Installation & Setup
Clone the repository:
```bash
git clone https://github.com/nfs-tech-bd/subgenAI
cd subgenAI
```
### Create and activate a virtual environment (recommended)
**Windows**:
```bash
python -m venv venv
.\env\Scripts\ctivate
```
**macOS / Linux**:
```bash
python3 -m venv venv
source venv/bin/activate
```
Install dependencies:
```bash
pip install -r requirements.txt
```
⚠️ The first time you run the script, Whisper will download the chosen model (default = `base`).
---
## ▶️ Usage
### Basic (Automatic Output Name)
```bash
python subgenai.py --video "my_video.mp4"
```
👉 Output: `my_video.srt`
### Specify Output File
```bash
python subgenai.py --video "my_video.mp4" --output "custom_subs.srt"
```
### Choose a Model
```bash
python subgenai.py --video "my_video.mp4" --output "custom_subs.srt" --model "medium"
```
---
## 📊 Model Trade-offs
| Model | Speed 🚀 | Accuracy 🎯 | Size 💾 |
|---------|----------|-------------|---------|
| tiny | Fastest | Lowest | ~75 MB |
| base | Fast | Medium | ~142 MB |
| small | Balanced | Good | ~466 MB |
| medium | Slower | Very Good | ~1.5 GB |
| large | Slowest | Best | ~3 GB |
---
## 📜 License
This project is licensed under the **MIT License**.
---
💡 *Built with ❤️ by Nafis Fuad with python and a lot of curiosity.*