Text-to-video generation is a fast-growing field of artificial intelligence that converts natural language descriptions into photorealistic, temporally coherent videos. With high potential across applications like virtual reality, education, and entertainment, the field is in urgent need of solution to the present challenges of motion consistency, semantic alignment, and video length generation. In our proposed hybrid architecture, we couple diffusion models and Generative Adversarial Networks (GANs) to mitigate the current motion consistency, semantic alignment, and video length generation challenges. Through temporal attention mechanisms, we sustain motion continuity over frames and ensure gradual training that can deal with more complex and longer video lengths. The model also utilizes memory-efficient sampling and frame interpolation for sharper visuals, better motion dynamics, and longer video length with minimal computational burden. Real-world applications span immersive VR experiences, automated education content generation, and personalized media content generation. Addressing ethical issues like authenticity confirmation and bias minimization also promotes responsible AI deployment. This contribution paves the way for future breakthroughs in generative AI and the redefinition of digital content generation.