Google Unveils Imagen AI to Generate Images From Text
Google introduces Imagen, a new artificial intelligence system that creates detailed images from written descriptions. The new tool reportedly outperforms competing models like DALL-E 2 by utilizing advanced Transformer and diffusion technologies.
Google LLC introduces Imagen, an artificial intelligence system that automatically generates images based on text prompts provided by users. The tech giant claims this new system outperforms other sophisticated models in the category, including the highly regarded DALL-E 2 from OpenAI that made headlines earlier this year.
The Imagen system operates using two separate neural networks working together. The first network is a Transformer model, a natural language processing algorithm originally invented by Google in 2017, which analyzes the user's text description and converts it into a mathematical representation known as an embedding that the system can process.
After the text is translated into an embedding, a second neural network takes over to actually draw the image. This second component is a diffusion model, a type of AI trained by learning to remove visual errors known as Gaussian noise from pictures, allowing it to construct highly accurate images from the mathematical data provided by the first network.