Meta Unveils Make-A-Scene AI That Combines Sketches and Text for Image Generation
Meta introduces Make-A-Scene, a new generative AI system that allows artists to control image creation using both text prompts and freeform sketches. The tool solves the randomness of text-only AI by letting users dictate exact scene layouts.
Meta Platforms Inc. unveils an advanced generative artificial intelligence system called Make-A-Scene that empowers artists to bring their imaginations to life. By combining text descriptions with freeform sketches, users give the AI a precise blueprint to generate stunning visual representations of their ideas.
Existing generative AI systems rely primarily on text prompts, which leads to unpredictable and random results regarding an image's layout and composition. Make-A-Scene solves this issue by allowing creators to dictate the exact placement, size, and orientation of elements through simple drawings alongside their written prompts.
The system learns the relationship between visuals and text by training on millions of example images, using a novel method to capture the scene layout from the user's rough sketch. It then fills in the intricate details based on the accompanying text, giving artists unprecedented control over the final digital artwork.