turn any text to an image with google’s latest AI tool ‘imagen’

IMAGE BY GOOGLE RESEARCH CREATES PHOTOREALISTIC IMAGES FROM INPUT TEXT

 

Google Research has unveiled Imagen, a new text-to-image AI. Basically, the system can create photorealistic images from input text. 

 

‘We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding,’ says the official paper. ‘Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation.’

 

‘AN UNPRECEDENTED DEGREE OF PHOTOREALISM’

 

Google claims Imagen features an unprecedented degree of photorealism and a deep level of language understanding that surpasses its competitors. For it to work, the program takes texts — let’s say,‘Three spheres made of glass falling into the ocean. Water is splashing. Sun is setting. — and turns it into an image portraying exactly that. The resulting images can be either photorealistic or more of an artistic interpretation.

 

Imagen is not available to the public, but Google has shared several examples of how the AI works. For the project, Google created a comprehensive and challenging benchmark for text-to-image model called DrawBench. With it, the company can compare Imagen with other AI methods like VQ-GAN+CLIP, Latent Diffusion Models, and DALL-E 2. DrawBench found out humans prefer Imagen over the competitors.

 

To learn more about Imagen, go here where Google explains the project in depth.

 

project info:

 

name: Imagen

created by: Google Research

 

juliana neira I designboom

may 25, 2022