Native image editing is powered by Gemini 2.0 Flash
The feature can edit both AI-generated images and users’ images
Gemini also allows conversational multi-step image editing
Native image editing in the Gemini app is available in more than 45 languages globally
Photo Credit: Google
Google is expanding the artificial intelligence (AI)-powered native image editing feature to the Gemini app. Announced on Wednesday, this new capability will allow users to make specific changes to images without leaving the AI chatbot. This capability was first unveiled in March within the Google AI Studio platform. Powered by the Gemini 2.0 Flash AI model, the feature allows users to conversationally ask the chatbot to change elements, add, replace or remove objects, enhance the background, and more.
Gemini App Now Supports Native Image Editing
Gemini was fairly late to the image generation space compared to rivals such as ChatGPT, Copilot, and Midjourney. While the feature was introduced in early 2024, the Mountain View-based tech giant had to pause and roll back the capability after several users found the chatbot to be generating inaccurate and offensive images. Google then re-added image generation in Gemini in December 2024 with Gemini 2.0 Flash.
In a blog post, the tech giant announced that it is rolling out native image editing via Gemini to the Gemini app. The capability is being rolled out in a phased manner, and it can take a few days before all users have access to it. Notably, native image editing will be available to users globally in more than 45 languages.
The Gemini 2.0 Flash-powered feature was first unveiled in the AI Studio, but now it is being made widely available. It will allow users to either take an AI-generated image or upload a real photograph and conversationally make edits to it. For instance, a user can upload an image of a table and ask Gemini to add some flowers on top of it.
As per the post, users can prompt Gemini to make complex edits, such as changing the background, replacing objects, adding elements, and even making minor changes to the subject. An example of the latter would be to upload an image of yourself and ask the AI to show what you would look like with a different hair colour.
One risk with this feature is the rise of deepfakes. Google says its models are trained to refuse any offensive or potentially harmful requests. Additionally, images created or edited with native image generation will include the invisible SynthID digital watermark. The company is also testing adding a visible watermark on all AI-generated images.
Apple Challenges 'Unreasonable' EU Order to Open Up to Rivals3 June 2025
Search
