NOTE: For lipsynced video results, scroll below

1. Specify the file paths, and make a new conda environment with Python = 3.6

2. Now Install the necessary Liabraries from requirements.txt, and make a media folder, inside of which add the video and audio

3. Simply Run "Python main.py" and your Final video will be ready

Objectives Achieved:

1. Visual and Audio Quality Lip Sync: The project successfully lip-syncs videos with improved visual and audio quality, ensuring that the lip movements accurately match the spoken words.
2. Robustness for Any Video: Unlike the original Wav2Lip model, the developed AI can handle videos with or without a face in each frame, making it more versatile and error-free.
3. Support for Longer Videos: The model overcomes the limitations of the original Wav2Lip GAN model, now effectively lip-syncing longer videos exceeding 1 minute in duration.
4. Particular Segments Can be extracted easily, unlike the original Model, any part with or without face can be extracted, with te desired audio combined.

Metrics: I haven't trained the model of any further dataset, so the Metrics is the same for the model

1. Average Mean Squared Error = 5.050382572478908
2. Average Peak Signal to Noise Ratio = 40.32044758489997

Results include LipSync Videos for:

1. Hindi Voice-Over on English Video.
2. Long Videos with some No-face or Other than Target speaker face with a lot of head and hands movement & Telugu to Hindi Translation Voice-Over synced
   results can be reproduced using the colab notebook or can be accessed at this google drive for reference: https://drive.google.com/file/d/1zjxMi1p3S9SL9UuatoC-2RgWbepyZWUY/view?usp=sharing

NOTE:

1. Instead of rendering the whole video at once, my approach breaks it into small pieces, which makes the process faster.
   
2. I had a different approach of this, rather than just skipping the frames, with no face(Recommende), I wrote the code for extracting the videos with faces and without faces, and individually rendering them. Through this, I didn't made any changes int the internal filing of the model.
   
3. All my work is done in main.py