Google has introduced a new AI model for image editing, “Gemini 2.5 Flash Image,” now available to users in the Gemini app, as well as to developers through the Gemini API, Google AI Studio, and Vertex AI. This model allows for much more precise photo alterations based on text prompts, while maintaining the recognizability of people, animals, and details, which often pose challenges for competitors. For example, users can change clothing colors or combine multiple objects in a photo, and the result looks natural—without distorted faces or backgrounds.
The unique feature of “Gemini 2.5 Flash Image” is character stability: the model can maintain the same appearance of a person or object across different scenes, regardless of changes in pose, lighting, or environment. This opens up possibilities for creating image series or product catalogs where consistency and recognizability are important.
Google emphasizes that the model works with “multi-turn” dialogues, allowing users to make changes step-by-step and refine requests to the AI. Other features include combining multiple photos, transferring style or texture from one object to another, and creating complex compositions based on text prompts.
For safety, Google has added both a visible watermark and a digital marker SynthID in the metadata of each generated image to complicate the spread of fake photos. The company emphasizes that the models limit the creation of prohibited images, and users gain more control over the creative process but cannot violate platform rules.
