I’m sure quite a few people have been using it like crazy lately. Google Gemini of Nano Banana Make all kinds of images, right? We’ve written a lot of related tutorials, and Ada also made a very detailed tutorial video a while back. But even back then, I felt that although Nano Banana performed quite close to perfection, one of its biggest issues was when creating images.Will be affected by the aspect ratio of the original image.If you want to create an image with a specified aspect ratio, you have to manually feed in the image yourself, which is a bit inconvenient. However, according to the latest news on the Google Developers Blog, this problem has now been solved.
Google Gemini Nano Banana adds new features such as image aspect ratio settings.
According to Latest news from the Google Developers Blog, this time Gemini 2.5 Flash Image Added a new feature to adjust image ratios for content creators (currently limited to Google AI Studio and Vertex AI platform). It now supports 10 different aspect ratiosFrom the 21:9 cinematic aspect ratio of ultrawide screens, to the 9:16 vertical ratio for social media, and even support for a variety of common square and near-square ratios, it covers nearly all commonly used image aspect ratios.
It’s also easy to use: just use AI Studio to select Gemini 2.5 Flash Image (nano banana) to create images, and on the right side of the screen you’ll see an additional scale setting. The default value is Auto (which should be the same ratio as the original image):
The original screen had no scale setting:
Currently there are ten image aspect ratios to choose from, including 1:1, 9:16, 16:9… Whether you’re making YouTube cover banner images or vertical short video covers for TikTok and Instagram Reels, it has almost everything you need:
For example, if you now upload a 1:1 image to generate a three-view and set the aspect ratio to 16:9, the output will match the ratio you requested. This makes things much more convenient for people who frequently use Gemini for image generation, since there’s no need to pre-process an image to a specific ratio to give the AI as a reference.
But if you use nano banana in Gemini to generate images, specifying the image aspect ratio in the prompt won’t do anything, so if you want to create images, it’s better to go to AI Studio:
However, the bigger new issue with Gemini’s image generation lately is that the AI will just slack off—you often have to prompt it N times to get a single image, and sometimes it outright refuses to do it no matter what. This is rather troublesome, and I hope Google addresses it soon (which should be a common flaw of large models being lazy).
Beyond the new image aspect ratio feature, this official article also mentions that the impact of Gemini 2.5 Flash Image is already evident in startup teams and the game development community. It also provides two case studies that clearly illustrate how it is redefining the possibilities of image creation.
Cartwheel: Breaking Through the “Slot Machine Experience” to Let Artists Truly Take Control of Creation
Cartwheel has long been dedicated to improving the “randomness” of AI image generation, believing that artists should be able to control character poses and presentation with greater precision. They built the “Pose Mode” feature and tried integrating it with various generative models, but previous systems always struggled to balance character consistency with the model’s world knowledge. It wasn’t until Gemini 2.5 Flash Image emerged that Cartwheel found the answer. By combining their in-house 3D pose tools with this model, they successfully created a new system that can generate characters from any camera angle while maintaining both motion fidelity and visual realism.
Cartwheel Co-founder Andrew Carr Point out:
「其他模型無法在保持姿勢準確性的同時，兼顧完整的世界知識。但 Gemini 2.5 Flash Image 做到了。」
Volley: Real-Time Game Visual Generation, Powering Interactive Entertainment
Another remarkable application comes from VolleyThis company developed an AI-driven dungeon exploration game. 《Wit’s End》Through Gemini 2.5 Flash Image, they can generate character portraits, scene visuals, and even multi-character composite images in real time during gameplay, with support for rapid multi-round editing.
Volley’s CTO James Wilsterman Expression:
「這個模型在遵循美學指引的同時，延遲依然維持在 10 秒以內，使得即時應用成為可能。玩家甚至可以透過語音或文字，反覆迭代出想要的風格。」
This capability not only greatly shortens the iteration time for art design, but also brings gamers an unprecedented immersive experience.
For developers, Google also provides detailed development documentation and sample programs, and at Google AI Studio It features a built-in “Build Mode” that lets users instantly generate AI applications by simply entering a prompt.
For example, entering “create an image editing app with filters” lets AI Studio directly generate a prototype and allows developers to deploy it or sync it to GitHub. This “zero-friction” creative workflow is especially friendly to startup teams and independent developers.
The official even launched a series of demo apps, such as:
-
BananimateUse the Nano Banana character to generate animated GIFs from images and prompts.
-
EnhanceAn inventive image magnifier with “infinite zoom” capability, plus a hidden easter egg (hint: 🍌).
-
Fit CheckUpload your personal photos and clothing to generate a virtual fitting room with try-on effects.
Source: KOCPC Chinese