Google Gemini's Nano Banana Image Model Gets Precision Upgrade

Google has rolled out a significant upgrade to its Gemini AI image model, codenamed Nano Banana 2, which allows users to make highly precise photo edits through natural language prompts while maintaining visual consistency in faces and objects—a capability lacking in competing tools like OpenAI's DALL-E.

The TechCrunch report details that Gemini 2.5 Flash Image, now available to all users in the Gemini app and via APIs, excels at tasks such as changing a subject's clothing color without distorting facial features, a common flaw in rival systems.

Google DeepMind product lead Nicole Brichtova emphasized the model's advancements in an interview, stating it "pushes visual quality forward" and "makes edits more seamlessly" compared to previous iterations, with benchmarks showing state-of-the-art performance on platforms like LMArena.

The new feature integrates with Google's Personal Intelligence system, which uses contextual data from connected Google services like Gmail and Google Photos to generate personalized images without requiring manual uploads or detailed descriptions. Users can now simply request "Create a claymation image of my family enjoying our favorite activity" and Gemini will automatically incorporate labeled photos from their Google Photos library.

Google clarified in its official blog post that the system does not directly train on private photo libraries but uses "limited info, like specific prompts in Gemini and the model's responses" to improve functionality, with privacy controls remaining opt-in via user settings.

This update follows the popularity of the original Nano Banana model, which overloaded Google's infrastructure in early 2025 due to overwhelming demand for personalized photo generation. As noted in a CNBC report, the feature propelled Gemini to the No. 1 spot on the Apple App Store, dethroning OpenAI's ChatGPT.

Google's move positions Gemini to compete more effectively with OpenAI's image tools, which drove ChatGPT's 700 million weekly users after GPT-4o's launch. Meta's recent partnership with Midjourney and Black Forest Labs' benchmark dominance further highlight the intensifying AI image generation race, where precise editing capabilities are becoming a critical differentiator.

The feature will roll out to paid subscribers (AI Plus, Pro, and Ultra tiers) in the United States within days, with broader availability planned for Chrome desktops and additional regions, according to the CNBC report. Google's approach—prioritizing user context over manual input—reflects a strategic shift toward embedding AI deeply within users' existing digital ecosystems rather than treating it as a standalone tool.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn