Google DeepMind Launches Nano Banana 2 Lite for Faster Image Creation
Google DeepMind 推出 Nano Banana 2 Lite，加速圖像生成
Introduction
Google has added Nano Banana 2 Lite, a faster image-generation model from the Gemini 3.1 family, to its tools for both developers and general users.
Google 將 Gemini 3.1 系列中速度更快的圖像生成模型 Nano Banana 2 Lite，加入到為開發者及一般使用者提供的工具中。
Main Body
The launch of Nano Banana 2 Lite shows a clear shift toward improving speed and reducing costs. While earlier versions, such as Nano Banana 2 and Nano Banana Pro, focused on creating highly realistic images and complex text, the Lite version is designed for quick testing and high-volume work. For example, it can generate an image in about four seconds, whereas the older Nano Banana model took twenty seconds.
Nano Banana 2 Lite 的推出顯示出提升速度與降低成本的明顯轉向。雖然早先的版本（如 Nano Banana 2 和 Nano Banana Pro）專注於創造高度擬真的圖像與複雜文字，但 Lite 版本旨在用於快速測試與高量工作。例如，它僅需約四秒即可生成一張圖像，而舊款的 Nano Banana 模型則需花費二十秒。
Cost is another major advantage of this new model. The API price is set at $0.034 per 1,000 images, which is much cheaper than the Pro version. However, this efficiency comes with some trade-offs in quality. Google admitted that the model sometimes makes mistakes in infographics, produces blurry text at small sizes, and struggles to keep characters looking the same in different images. Despite these issues, user ratings from Arena.ai suggest that people find the quality very similar to the non-Lite versions.
成本是此新模型的另一個主要優勢。API 價格設定為每 1,000 張圖像 0.034 美元，比 Pro 版本便宜許多。然而，這種效率在品質上有所妥協。Google 承認該模型有時在資訊圖表上會出錯，在小尺寸下會產生模糊文字，且難以在不同圖像中保持角色外觀一致。儘管存在這些問題，Arena.ai 的用戶評分顯示，使用者認為其品質與非 Lite 版本非常相似。
At the same time, Google has expanded the availability of Gemini Omni Flash for video generation and introduced Omni Product Studio to turn still images into e-commerce videos. These moves are part of a larger trend of using AI in creative fields, highlighted by Google's $75 million deal with the film company A24. To prevent the spread of fake media, Google has added SynthID watermarks to all images created by Nano Banana 2 Lite.
同時，Google 擴展了 Gemini Omni Flash 在影片生成方面的可用性，並推出 Omni Product Studio 將靜態圖像轉化為電商影片。這些舉措是 AI 應用於創意領域大趨勢的一環，其中最顯著的是 Google 與電影公司 A24 達成的 7,500 萬美元協議。為了防止偽造媒體的傳播，Google 已為所有由 Nano Banana 2 Lite 生成的圖像添加了 SynthID 浮水印。
Conclusion
Nano Banana 2 Lite is now available through Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform, replacing the original Nano Banana model.
Nano Banana 2 Lite 現在可透過 Google AI Studio、Gemini API 及 Gemini Enterprise Agent Platform 使用，取代了原有的 Nano Banana 模型。
Vocabulary Learning
⚡ The Power of Comparison
To move from A2 to B2, you must stop using simple sentences (like "It is fast. It is cheap") and start connecting ideas to show contrast.
Look at this goldmine from the text:
"...it can generate an image in about four seconds, whereas the older Nano Banana model took twenty seconds."
The B2 Secret: "Whereas" In A2, you probably use "but." In B2, we use whereas to balance two different facts in one elegant sentence. It acts like a scale, weighing two opposite realities.
How to use it like a pro:
- A2 Style: The Pro version is expensive. The Lite version is cheap.
- B2 Style: The Pro version is expensive, whereas the Lite version is much more affordable.
🛠 Vocabulary Shift: "Trade-offs"
Notice the phrase: "this efficiency comes with some trade-offs in quality."
At a B2 level, you need words that describe complex situations. A "trade-off" is when you accept something bad (lower quality) to get something good (higher speed).
Example for your life: "Living in a small apartment is a trade-off; I have less space, but I save a lot of money on rent."
🔍 Spotting the Nuance: "Struggles to..."
Instead of saying "The AI cannot do this," the author says the AI "struggles to keep characters looking the same."
Why this matters: B2 speakers use "softeners." Saying something "struggles" is more precise than saying it "can't." It means the AI tries, but it isn't perfect. Use this when you want to describe a difficulty without being too absolute.