Google has launched Imagen 3, described as the company’s highest-quality text-based image generation AI model yet. Imagen 3 would be able to generate images with better detail, richer lighting and fewer artifacts than previous versions.
Google also boasts that it has improved the Imagen 3 AI model’s ability to understand text instructions, helping it generate images in a variety of visual styles, being able to understand minor details from more complex text instructions.
The new AI model is capable of generating high-quality images in a variety of formats and visual styles, from sketches to photorealistic landscapes and even emulating the characteristics of an oil painting or plasticine creations.
Google says that Imagen 3 understands written text instructions in a natural way, using everyday language, making it easier to use than alternatives that require the use of certain exact phrases to achieve the desired results.
More details, visual styles, but also increased security
The extensive databases with which Imagen 3 was trained allows the model to understand concepts such as the specific camera angles through which the generated image should be rendered, with much richer detail and much improved lighting. For example, the image generator can add details such as wrinkles specific to human skin or complex textures such as the fabric of a fabric.
Also, to limit the possibility of Imagen 3 generating images with content that could be considered offensive or harmful, the AI model has been equipped with numerous extensive filters as well as the latest privacy, safety and security technologies. Among these, SynthID stands out, an advanced watermark placement tool, which includes one such digital identification element in the pixels of each image. This is imperceptible to the human eye, but easy to detect in case of need.
You can access Imagen 3 here and you can test it through Gemini by clicking on the “Try it on Gemini” button.
Source: Google
Source: www.go4it.ro
