Google has rolled out a significant upgrade to the image editing capabilities of its Gemini AI, powered by the new Gemini 2.5 Flash Image model. This update allows users to edit photos with natural language prompts directly within the Gemini app. The AI model maintains the original composition, objects, and people while adding, changing, or removing elements as instructed. Users can now make iterative modifications, blend subjects from different photos, and change backgrounds, all while preserving the likeness of individuals.
DeepMind's Gemini 2.5 Flash Image, previously known as 'Nano Banana' during testing, outperformed other photo-editing models, gaining recognition as a top-rated editing model for its consistency, quality, and instruction following. The model is accessible through the Gemini API, Google AI Studio, and Vertex AI for developers. Adobe also plans to integrate the new model into its Firefly app and Adobe Express. All images created or edited with Gemini 2.5 Flash Image will include an invisible SynthID digital watermark.
This upgrade focuses on providing more precise edits and reducing errors, enabling users to achieve advanced editing effects such as costume changes, location swaps, and stylistic remixes. The Gemini 2.5 Flash Image model is priced at $30.00 per 1 million output tokens, with each image costing approximately $0.039.
