Subtitle Generator

This project automates subtitle generation for videos by:

• Extracting audio from a video
• Transcribing and translating speech using OpenAI's Whisper
• Generating SRT subtitle files with accurate timestamps
• Optionally burning subtitles into the video

Features

✅ Multilingual Support – Detects and translates speech to English
✅ Automatic Time Syncing – Generates SRT subtitles with timestamps
✅ Fast Processing – Uses Torch & Hugging Face Transformers for efficient speech recognition
✅ GPU Support – Runs on CUDA for faster transcription (if available)

---

Installation

1. Install Dependencies

For GPU Users (CUDA 12.6)

pip install -r requirements-gpu.txt

For CPU Users

pip install -r requirements-cpu.txt

2. Install FFmpeg

If FFmpeg is not installed, install it via:

• Windows: Download FFmpeg
• Linux/macOS: Run:

  sudo apt install ffmpeg  # Ubuntu/Debian
  brew install ffmpeg  # macOS

---

Usage

Run the subtitle_generator.ipynb Jupyter Notebook to process a video and generate subtitles.

1. Extract Audio from Video

import subprocess

video_path = "/path/to/video.mp4"
audio_path = "audio.wav"

subprocess.run(["ffmpeg", "-i", video_path, "-q:a", "0", "-map", "a", audio_path, "-y"])
print("Audio extracted successfully!")

2. Transcribe & Translate Speech

from transformers import WhisperProcessor, WhisperForConditionalGeneration
import torch, librosa

# Load Whisper Model
model_name = "openai/whisper-large-v2"
processor = WhisperProcessor.from_pretrained(model_name)
model = WhisperForConditionalGeneration.from_pretrained(model_name).to("cuda" if torch.cuda.is_available() else "cpu")

# Load & Process Audio
audio, sr = librosa.load(audio_path, sr=16000)
inputs = processor(audio, sampling_rate=16000, return_tensors="pt").input_features.to("cuda" if torch.cuda.is_available() else "cpu")

# Transcribe & Translate
forced_decoder_ids = processor.get_decoder_prompt_ids(language="en", task="translate")
with torch.no_grad():
    result = model.generate(inputs, forced_decoder_ids=forced_decoder_ids)

transcription = processor.batch_decode(result, skip_special_tokens=True)[0]
print("Transcription:", transcription)

3. Generate SRT Subtitle File

from datetime import timedelta

srt_path = "subtitles.srt"

def format_timestamp(seconds):
    return str(timedelta(seconds=int(seconds))).replace(".", ",") + ",000"

with open(srt_path, "w", encoding="utf-8") as srt_file:
    for i, segment in enumerate(transcription.split(". ")):  
        start_time = i * 5  
        end_time = start_time + 5  

        srt_file.write(f"{i+1}\n{format_timestamp(start_time)} --> {format_timestamp(end_time)}\n{segment}\n\n")

print(f"Subtitles saved to {srt_path}")

4. Burn Subtitles into Video (Optional)

ffmpeg -y -i /path/to/video.mp4 -vf subtitles=subtitles.srt -c:a copy output.mp4

---

Folder Structure

/subtitle_generator/
│── subtitle_generator.ipynb  # Jupyter Notebook for processing
│── requirements-gpu.txt      # Dependencies for GPU users
│── requirements-cpu.txt      # Dependencies for CPU users
│── subtitles.srt             # Generated subtitle file
│── output.mp4                # Video with embedded subtitles (optional)

---

License

This project is open-source under the MIT License.
Feel free to use, modify, and contribute! 🎉

---

Author

👨‍💻 aakash j.v.v
📧 aakashjammula6@gmail.com
🌐 linkedin