app.py

# Install dependencies if not already done in your environment
# pip install transformers torch gtts ffmpeg-python

from transformers import pipeline
from gtts import gTTS
import subprocess

# Step 1: Extract Text from Audio using Hugging Face Transformers
def extract_text_from_audio(audio_path):
    # Load the ASR pipeline from Hugging Face with a Whisper-like model
    transcriber = pipeline("automatic-speech-recognition", model="openai/whisper-base")
    
    # Transcribe the audio file
    transcription = transcriber(audio_path)
    text = transcription["text"]
    
    # Save transcribed text to a file (optional)
    with open("video_text.txt", "w") as f:
        f.write(text)
    
    return text

# Step 2: Generate Voice-Over using gTTS
def generate_voice_over(text, output_audio_path="voice_over.mp3"):
    # Generate audio with gTTS
    tts = gTTS(text=text, lang="en")
    tts.save(output_audio_path)
    print(f"Voice-over saved as {output_audio_path}")
    return output_audio_path

# Step 3: Combine Voice-Over with Original Video using FFmpeg
def add_voice_over_to_video(video_path, audio_path, output_video_path="output_video_with_voice.mp4"):
    # Use FFmpeg to combine video with new audio
    ffmpeg_command = [
        "ffmpeg",
        "-i", video_path,
        "-i", audio_path,
        "-c:v", "copy",
        "-map", "0:v:0",
        "-map", "1:a:0",
        "-shortest",
        output_video_path
    ]
    subprocess.run(ffmpeg_command)
    print(f"Final video with voice-over saved as {output_video_path}")

# Run the complete process
def main(video_path):
    # Step 1: Extract text from video/audio
    text = extract_text_from_audio(video_path)
    print("Extracted Text:", text)
    
    # Step 2: Generate voice-over from extracted text
    audio_path = generate_voice_over(text)
    
    # Step 3: Add voice-over to the video
    add_voice_over_to_video(video_path, audio_path)

# Provide the path to your input video file
main("input_video.mp4")