Google also announced Imagen-3 which is the latest version of its image generation model at the I/O development event on Tuesday. It is enhanced with the ability to generate images with more realistic and accurate visuals. Details are also produced in more detail with better lighting and composition. This includes small details like the fine wrinkles in a human hand, and complex textures like a large bouquet of colorful flowers in a vase.
According to Google, Imagen-3 can also handle text better than the previous model. If the prom asks for some text to be inserted into the image it wants to generate, it can be done exactly as compared to the previous model. More interesting is that the Imagen-3 is smarter to understand the details than the long prom. Then each generated image also has a SynthID watermark that makes it easy to trace its origin.
Also informed that editing features such as inspainting and outpanding will also be available in Imagen-3 in the coming months. Starting today, Imagen 3 is available to select creators in private preview through ImageFX. Imagen 3 is also said to be coming soon to Vertex AI.