Imagen by Google

Imagen is a text-to-image technology created by Google Research that enables the generation of photorealistic images from a text description. It combines the power of a large transformer language model to understand the text and the strength of diffusion models to produce high-quality images. Research has shown that generic large language models (e.g. T5), pretrained on text-only datasets, are effective at encoding text for image synthesis. Increasing the size of the language model in Imagen significantly improves both sample quality and image-text alignment more than increasing the size of the image diffusion model.