谷歌 AI 图像模型 Nano Banana 2.1 发布,提升视觉设计、蒙版编辑、主体一致性Google Releases AI Image Model Nano Banana 2.1, Improving Visual Design, Mask Editing, and Subject Consistency


AI free token/AI免费积分
感谢IT之家网友 软媒使用者1053564、咩咩洋、愚公骑马、华南吴彦祖、补药吖、Domado 的线索投递!

IT之家 10 月 7 日消息,谷歌于今日正式发布了 Nano Banana 2.1,这是该公司 AI 图像生成与编辑模型的升级版本。谷歌强调,新版 Nano Banana 2.1 达成了全方位提高,在视觉外观、基于蒙版的图像编辑以及主体一致性这几维度的进步尤为突出。

谷歌 AI 图像模型 Nano Banana 2.1 发布,提升视觉设计、蒙版编辑、主体一致性

今年早些时候,谷歌推出了 Nano Banana 2,也被称作 Gemini 3.1 Flash Image。Nano Banana 2 在具备谷歌 Flash 系列模型高速特性的同时,达成了与 Nano Banana Pro 同级别的出图质量。谷歌还把多项原本属于 Pro 版本的能力下放至这款定位更低的模型,涵盖现实世界知识库、文字渲染成效优化,还有角色与物体的一致性表现提高。

IT之家注意到,Nano Banana 2.1 在原有能力的基础之上,主要在三大方向做出改进:

改进后的主体一致性对于 Gemini 使用者而言实用性很强,可以保证同一个人物或者物体在多次生成结果当中保持样貌统一。谷歌 Nano Banana 2.1 现已逐步部署上线,覆盖 Gemini 应用、谷歌搜索的 AI 模式、Google AI Studio、Google Flow、Stitch、谷歌广告以及 Gemini 企业平台。

面向开发者,Nano Banana 2.1 已经正式开放,对应的模型 ID 为 gemini‑nano‑banana‑2.1;该模型兼容文本、图片、音频以及视频输入,同时可以输出 1K、2K 以及 4K 分辨率的图像。

当下,OpenAI 的 ChatGPT Images 2.5 依旧是全球领先的图像生成编辑模型。在使用者持续微调一张图片时,它更擅长保留上一轮的修改成果,不会改动没有编辑的区域,也不会随着反复迭代让原图画质变差。紧随 OpenAI 这款机型之后的是微软 AI 的 MAI‑Image‑2.6 模型,该模型于今年 8 月发布。

本文来源:https://www.ithome.com/1/010/149.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。

Thanks to IT Home readers Soft Media User 1053564, Miemiyang, Yugong Qima, Huannan Wu Yanzu, Buyaoa, and Domado for submitting the tip!

IT Home, October 7 — Google officially released Nano Banana 2.1 today, an upgraded version of its AI image generation and editing model. Google emphasized that the new version achieves comprehensive improvements, with particularly notable advances in visual appearance, mask-based image editing, and subject consistency.

Google Releases AI Image Model Nano Banana 2.1, Improving Visual Design, Mask Editing, and Subject Consistency

Earlier this year, Google launched Nano Banana 2, also known as Gemini 3.1 Flash Image. Nano Banana 2 combines the high-speed capabilities of Google's Flash series models with image generation quality on par with Nano Banana Pro. Google has also brought down multiple Pro-version features to this lower-tier model, including real-world knowledge base integration, optimized text rendering, and improved consistency for characters and objects.

IT Home notes that Nano Banana 2.1 primarily improves in three areas beyond its original capabilities:

The improved subject consistency is highly practical for Gemini users, ensuring that the same person or object maintains a consistent appearance across multiple generations. Google Nano Banana 2.1 is now gradually rolling out, covering the Gemini app, Google Search AI mode, Google AI Studio, Google Flow, Stitch, Google Ads, and the Gemini enterprise platform.

For developers, Nano Banana 2.1 is officially available with model ID gemini-nano-banana-2.1; it supports text, image, audio, and video input, and can output images at 1K, 2K, and 4K resolutions.

Currently, OpenAI's ChatGPT Images 2.5 remains the world's leading image generation and editing model. When users continuously refine a single image, it excels at preserving previous edits, leaving unedited areas untouched, and avoiding quality degradation through repeated iterations. Following OpenAI's model is Microsoft AI's MAI-Image-2.6, released this past August.

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry