KevinGeLe / SmartSRT

📄 SmartSRT is a command-line tool for generating accurate subtitles with per-word timestamps. It uses WhisperAI for speech transcription, NVIDIA NeMo for diarization, and OpenCV for face recognition. The program is good at creating high accuracy subtitles. 🎧💻⚙️

SmartSRT

SmartSRT is my attempt, at an community run end-to-end solution for automatically generating subtitles for videos with a maximum subtitle length constraint, along with speaker diarization. It uses machine learning models for speech recognition, text summarization, face regocnition, and diarization, and will run on a CUDA GPU for faster performance.

# Getting Started You need to install the required dependencies and download the necessary models.\ ⚠️Remember that you have to use a CUDA supported GPU!⚠️

Clone the repository:
git clone https://github.com/KevinGeLe/SmartSRT.git
Install the required dependencies:
cd SmartSRT
pip install -r requirements.txt
The necessary models will be downloaded, by specifying the models that you want to use.

TODO:

• Add Face recognition (±1 second)👱
• Add Max-Subtitle-Lenght with Per-Word-Timestamps:🕰️
• Add new parsers for Max-Lenght, Output, Input and Model🔍
• Refactor code to improve readability🐊
• Work on improving performance🦈
• Update README.md📑

License

SmartSRT is released under the MIT license. See the LICENSE file for more information.