Google is pushing a lot on artificial intelligence . In fact, soon there will be several projects of the company that will benefit from this technology, starting from the Gboard keyboard for Android devices.
Specifically, the most recent beta version of the application – version 12.7.05.507749191 – contains code strings that refer to an Imagen Keyboard.
This new option – which could be placed in the panel dedicated to keyboard shortcuts, such as GIF – should allow you to generate images from text input.
Imagen, announced last May, will combine a deep level of text understanding with an “unprecedented degree of photorealism”.
In a benchmark comparison last year and including VQ-GAN-CLIP, Latent Diffusion Models and DALL-E 2 , Google said that human users preferred
“Imagen over other models in side-by-side comparisons, both in terms of sample quality than the combination of images and text”.
In addition, the company stated that Imagen is also more efficient in spatial relationships, long texts, rare words and more challenging suggestions.
Finally, the company added that
“Our key finding is that large generic language models (eg T5), pre-trained on text-only corpora, are surprisingly effective at encoding text for image synthesis: the increased size of the language model in Imagen results in a dramatic increase in both sample fidelity and image-text alignment, far more than increasing the size of the image diffusion model.”