Googles Gemini AI Image Editor Hits and Misses
After all the buzz surrounding Google's newest generative AI tool, playfully nicknamed the "nano bananas" model, I decided to put it to the test. This tool, officially known as Gemini 2.5 Flash Image, upgrades your ability to edit photos directly within the Gemini interface. The "nano bananas" moniker gained traction after a series of themed teasers from Google executives.
The tool has seen massive adoption since its launch. According to Josh Woodward, Google's VP of Google Labs and Gemini, users have created over 200 million AI images, and the Gemini app has surpassed 10 million sign-ups.
Google has been heavily investing in its generative media models, showcasing updated image and video generators at its annual I/O conference. Their AI video generator, Veo 3, impressed with synchronized audio, and its AI filmmaker tool, Flow, has been used to create over 100 million videos. Having tested numerous AI creative tools, I was eager to see Google's latest offering. However, my experience with Gemini 2.5 Flash Image shows that a splashy debut doesn't always guarantee a flawless performance. Here’s a breakdown of my experience: the good, the bad, and the genuinely frustrating.
The Good: Where Gemini's AI Editor Shines
The Gemini "bananas" model is surprisingly proficient at adding new elements to your photos, blending them in a way that looks natural. It also does a great job maintaining character consistency, meaning the people in my photos didn't look distorted after AI processing. These are two significant challenges for AI image tools, and Google has clearly focused on improving them.
You can see this in action with a picture of my sister and me. I asked Gemini to add a third sister who looked like us, and it did a scarily convincing job, placing a new person right between us while keeping our original appearances intact.
Welcome to the family, scarily accurate-looking AI sister.
I was also impressed by the speed. Gemini consistently delivered edited images in under 15 seconds, which is excellent. Another positive feature is the automatic watermarking of all AI-edited images. While the use of the sparkles emoji for AI is debatable, having clear markers for AI-generated content is crucial.
While Gemini can also create AI images from scratch, I'd recommend using a more specialized AI image generator for that purpose, as they offer more fine-tuned controls to get the exact result you want.
The Bad: Frustrating Flaws and Limitations
Unfortunately, Gemini's image editor has some serious limitations. It automatically generates square images and completely ignored my follow-up prompts asking to change the dimensions.
A more significant issue was the reduction in photo resolution. I take most of my photos with an iPhone 16, known for its excellent cameras, but the Gemini model often blurred the fine details. This is a major drawback that will likely deter photographers.
Despite my request to brighten up the target's rings, the original photo (left) has more richness and texture than the AI-edited one (right).
Where the AI really struggled was with prompt adherence for complex edits—the very tasks where AI is supposed to shine by automating tedious work. For example, I repeatedly asked Gemini to remove reflections from a photo of a Freakier Friday movie poster, but it failed every time. With each attempt, the image quality worsened until the text was illegible and the faces of Lindsey Lohan and Jamie Lee Curtis were bizarrely distorted.
All I wanted was for Gemini to remove the reflection in the upper-right corner of the poster, but instead it warped the movie tag line and faces.
Google's Response and The Final Verdict
I reached out to Google about the resolution and dimension issues. A spokesperson confirmed the company is "aware and actively working on both issues. It's been a big update from our previous model but we'll continue to improve on the model."
Overall, the "nano bananas" model proves Google is serious about competing in generative media. However, it currently has significant pitfalls, focusing more on adding new elements than on improving existing photos with common edits. For now, this tool is best for Gemini fans who want to make big, creative changes quickly. If you're looking for precision and control, you'll have to wait for Google's next update or look for a different tool.
How to Access Gemini's Image Editor
The new model is automatically integrated into the base Gemini 2.5 Flash model. Gemini is available for free, with more advanced models and higher usage limits included in Google's AI plans, which start at $20 per month.
Paying subscribers might also find the model in Google AI Studio. Simply upload an image and type your prompt. According to Google's privacy policy for Gemini, the company may use uploaded information to improve its products, so it's best to avoid uploading sensitive data. Additionally, the AI prohibitive use policy bans the creation of illegal or abusive content.
