Meta has unveiled a new tool that provides artists more creative control over platforms like OpenAI’s Dall-E 2 and Google’s Imagen that use AI to generate images.
According to Meta, for AI to reach its full potential, users must be able to easily and intuitively shape the information that a system produces. Dubbed Make-A-Scene, Meta’s newest AI system is a multimodal generative technique that will allow users to use both text descriptions and freeform doodles to explain and illustrate their vision.
Make-A-Scene, like other generative AI models, trains on millions of example images to learn the link between visuals and text.
