Generative artificial intelligence (AI) has issues generating images consistently. The same prom produces different iterations, making the process of refining images difficult. Today Google launched Gemini 2.5 Flash which generates more consistent images because of the reference image feature.
For example, if a user wants to generate an image with their face, their own photo reference image can be given to Gemini 2.5 Flash. The result is an image generated with their own photo and no longer an unrecognizable caricature like in other generative AI models.
In addition to personal photos, background images and objects that they want to use can also be given to Gemini 2.5 Flash. In the demo, users can place a light object in the background image with the location where the light is placed can be set by the user.
Gemini 2.5 Flash is also capable of performing advanced image editing. For example, changing the color of the shirt, the style of the shirt or removing dirt from the shirt worn. The body style in the image, the background object and the color of the image can also be changed.
This latest model can be accessed using Gemini API and Google AI Studio for developers and Vertex AI for businesses.