Gemini improves AI image editing
Google is adding a new image-editing model to Gemini that better preserves the appearance of people and pets while changing clothes, backgrounds or settings. It also lets you combine photos, edit images in multiple steps and apply styles from one image to another, with marks indicating the use of AI.
Google is adding a new image-editing model to the Gemini app to address one of the most noticeable problems with these tools: keeping a person, pet or character looking like themselves after several changes.
The update has been available since August 26, 2025. You can upload a photo and ask Gemini to change the clothes, background, time period or even the location, without altering the original features too much.
Your appearance stays consistent during editing
The feature is designed to prevent an edit from turning you into someone else. For example, you can see what you would look like with a 1960s hairstyle, place your dog on a basketball court or change a roomâs walls while keeping the elements you do not want to modify intact.
Google says its evaluations rate the new model as the best in the world for image editing. The company also says it focused development on preserving the likeness of people, family members, friends and pets from one image to the next.
More control for creating an image step by step
Gemini now lets you upload several photos and combine them to create a new scene. You can combine a portrait of yourself with one of your pet, for example, and ask for an image showing both of you together in another location.
It also supports multi-step editing. That means you can start with an empty room, paint the walls and then add a bookcase, sofa or table without having to recreate the entire image each time.
Available options include:
- Change a personâs or petâs clothes, setting, apparent profession or time period.
- Combine several images into a new composition.
- Modify only part of an image while keeping the rest unchanged.
- Apply the style, color or texture of one image to an object in another, such as using a butterfly wing pattern to design a dress.
When you finish an image, you can also upload it to Gemini again to turn it into a video, Google explains.
Images include an AI watermark
All images created or edited in Gemini include a visible watermark. They also contain an invisible digital watermark called SynthID, designed to indicate that the image was generated or modified with artificial intelligence.
For you, the main change is practical: Gemini is no longer limited to generating images from scratch and now offers a more useful way to modify specific photos without completely losing their identity. The point to watch is whether that likeness remains reliable in complex scenes and after many consecutive edits.