Text-to-Video 3D

Text-to-Video Synthesis with VQGAN and CLIP

This project combines the capabilities of VQGAN (Vector Quantized Generative Adversarial Networks) and CLIP (Contrastive Language-Image Pre-training) to generate video frames based on textual descriptions. The synthesis system applies a series of transformations to create dynamic, artistic representations of textual prompts.

Features

• Text-to-image synthesis using VQGAN+CLIP.
• Image transformations (zoom, rotate, translate) to enhance the visual dynamics.
• Output frames saved as images which can be compiled into videos.

Prerequisites

• Python 3.8 or higher
• Pip package manager
• Access to a CUDA-compatible GPU for faster processing (optional but recommended).

Installation

1. Clone the repository

   git clone https://github.com/your-repository/Text-to-Video-VQGAN-CLIP.git
   cd Text-to-Video-VQGAN-CLIP

2. Install dependencies

   pip install -r requirements.txt

3. Install dependencies Download the necessary VQGAN model configuration and checkpoint files. These files should be placed in the appropriate directories specified in the main script.

Usage

Run the main script to generate frames based on predefined text prompts. Each frame is saved as an image in the output directory.

python src/main.py

You can modify the text prompts directly in the main.py file to create different images.

Extending the Project

Feel free to add more functionalities, such as:

• Real-time text input for generating images on the fly.
• Integration with web frameworks for an interactive user interface.
• More complex transformations and effects.