October 23, 2023: Meta, the parent company of Facebook and Instagram, announced the progress of its AI model called Imagen, which is capable of generating realistic images from text descriptions. 1
It said that the model was trained on a massive dataset of images and text, and it can create images in various styles, including photorealistic, artistic, and abstract.
Imagen is a significant advancement in the field of AI-generated imagery. Previous AI models could generate images, but they were often unrealistic or cartoonish.
Meta claims that this can generate indistinguishable images from real photographs, even for complex scenes with multiple objects and people.
Meta’s AI image generator employs a diffusion model to create realistic images from text descriptions. Unlike traditional AI image generators that rely on generative adversarial networks (GANs), Imagen adds and gradually removes noise from a blank canvas to construct recognizable images.
In a blog post announcing Imagen, Meta AI researchers said that the model could be used for various applications, including creating new forms of creative content, improving the quality of existing content, and developing new tools for education and research. 2
Meta AI researchers also said that Imagen is still under development, and they are working on making the model more efficient and accessible to a wider range of users.3
- Meta AI on X: https://shorturl.at/dgiBS[↩]
- Meta : https://ai.meta.com/blog/?page=1[↩]
- Meta AI on X: https://shorturl.at/boszK[↩]
