Meta Unveils Make-A-Video AI That Turns Text Prompts Into Moving Clips
Meta introduces Make-A-Video, a new artificial intelligence system that generates short videos from simple text descriptions. The tool builds on previous text-to-image technology by teaching the AI to understand real-world motion.
Meta CEO Mark Zuckerberg introduces Make-A-Video, a new artificial intelligence system that creates short videos directly from text prompts. The AI generates impressive moving images, such as a teddy bear painting a self-portrait or a spaceship landing on Mars, by utilizing deep learning from image synthesizers like GPT-3 and LAION-5B.
Generating video presents a much greater challenge than creating static images because the system must accurately predict how each pixel changes over time. Make-A-Video solves this complex problem by adding an unsupervised learning layer that helps the AI understand motion in the physical world and apply it to traditional text-to-image generation methods.
Meta plans to share a full demo of the technology in the future, though it remains unclear when the general public will get to try it. This announcement follows a massive surge in generative AI tools, coming just after DALL-E removes its waitlist and as competitors like Stable Diffusion and Midjourney continue to rapidly evolve.