Yesterday, Google expanded the number of countries in which its Gemini Pro multimedia model is integrated into Bard, its AI assistant. But also Announce the ability to create images directly in chat, with realistic results.
The foundation is the Imagen 2 engine, which is now integrated with Bard, ImageFX, Search and Vertex AI. But at the moment none of this is available for users in Portugal.
Imagen 2 is an image generation model, a competitor to OpenAI's DALL.E that powers Microsoft's ChatGPT and Copilot.produces realistic images with the help of well-targeted prompts, which also allows you to edit the image without having to repeat the process.
See some images created with Imagen 2 and prompts used
The quality of user prompts will make a difference in results, however The quality of detail and accuracy of the result can make the difference between image-trained models, making the connection between the text and the generated image. Faces and hands are now more realistic and have adapted aesthetics based on human preferences, with frames, lighting and details.
Images can also be edited during the process, using the In-Draw and Out-Draw modesWhich becomes useful for adding items. Some of these features won't be available until later in 2024.
Google has an Imagen API on Google Cloud Vertex AI for developers to explore features. You can learn more about this Google Images template using Documentation published on Vertex AI, detailing how it works and the creative possibilities.
“Friendly zombie fanatic. Analyst. Coffee buff. Professional music specialist. Communicator.”