Adobe camera app’s new feature will critique your photos using AI

Adobe 相机应用推出新功能：利用 AI 为你的照片提供点评

Adobe is adding new AI-powered features to its experimental iOS camera app called Project Indigo, launched last year. The app previously offered pro controls, multi-frame super-resolution, and different capture modes, and is now adding features that will use LLMs (large language models) to critique photos and provide editing suggestions. Adobe 正在为其去年推出的实验性 iOS 相机应用“Project Indigo”增加新的 AI 功能。该应用此前已提供专业控制、多帧超分辨率和多种拍摄模式，现在又新增了利用大语言模型（LLM）来点评照片并提供编辑建议的功能。

It’s also adding other AI features, like advanced object removal, depth of field generation, and the ability to add different styles to photos. Marc Levoy, the person heading Adobe’s project, had previously developed Pixel’s camera chops. He said that most generative AI tools provide prompt-based editing and, often, finding the perfect prompt to tweak a photo or get the right result can be a tricky endeavor. That’s why most of the new experimental AI features are buttons that can generate more deterministic outputs. 它还增加了其他 AI 功能，如高级对象移除、景深生成以及为照片添加不同风格的能力。负责该项目的 Marc Levoy 此前曾开发过 Pixel 的相机功能。他表示，大多数生成式 AI 工具都提供基于提示词（prompt）的编辑，但要找到完美的提示词来调整照片或获得理想结果往往是一项棘手的任务。因此，大多数新的实验性 AI 功能都设计为按钮形式，以生成更具确定性的输出结果。

For instance, Google’s camera coach feature for Pixel phones, launched last year, offered more generic framing suggestions. Project Indigo’s features, by comparison, are fairly descriptive and could help you learn some photography tricks, even if you don’t agree with the AI assessment. There are two features in this category. First is the photo critique, which includes a “professional” opinion about framing, lighting, colors, and emotional impact. 例如，谷歌去年为 Pixel 手机推出的“相机教练”（camera coach）功能提供的是较为通用的构图建议。相比之下，Project Indigo 的功能描述相当详细，即使你不同意 AI 的评估，它们也能帮助你学习一些摄影技巧。此类功能共有两项：第一项是照片点评，包含关于构图、光线、色彩和情感表达的“专业”意见。

The second is capture and edit suggestions, which give you tips about reshooting the photo on how you can change framing, exposure, and objects in the viewfinder. For instance, the app told me to remove the hexagonal white object in the frame from a photo I took. A second section tells you how you can use an existing photo to make it better using Adobe Lightroom controls. 第二项是拍摄和编辑建议，它会为你提供关于重拍照片的提示，告诉你如何改变构图、曝光以及取景器中的物体。例如，该应用曾建议我移除我所拍照片中画面里的六边形白色物体。另一部分则会告诉你如何利用现有的照片，通过 Adobe Lightroom 的控制功能对其进行优化。

Photography apps, including Apple Photos, Google Photos, and Adobe Photoshop, have offered object removal for years. But the feature often depends on the user circling or selecting an object by drawing on the screen, which is not perfect every time. Project Indigo’s new feature gives you toggles for things to remove from an image, like people in the background, trash and trash cans, wires and poles, fences, vehicles, and other clutter. You can also describe a custom object to remove. 包括 Apple Photos、Google Photos 和 Adobe Photoshop 在内的摄影应用多年来一直提供对象移除功能。但该功能通常依赖于用户在屏幕上圈选或绘制来选择对象，效果并不总是尽如人意。Project Indigo 的新功能为你提供了可移除对象的开关，例如背景中的人物、垃圾和垃圾桶、电线和电线杆、围栏、车辆以及其他杂物。你还可以描述一个自定义对象进行移除。

The results of this feature are pretty impressive. The app removed my friend and the object he was holding from the background without creating any strange artifacts. The app also allows you to use AI to create depth of field for a photo to simulate a blurred background. With this new update, Adobe is also experimenting with the style transfer feature, which lets you turn your picture into tones like watercolor, pen and ink, ink line with color wash, monochromatic, and backlit subject. 该功能的效果令人印象深刻。应用成功地将我的朋友和他手中的物体从背景中移除，且没有产生任何奇怪的伪影。该应用还允许你使用 AI 为照片创建景深，以模拟背景虚化效果。随着此次更新，Adobe 还在试验风格迁移功能，让你能将照片转换为水彩、钢笔画、水墨渲染、单色以及背光主体等风格。

Some styles remind me of the early days of the Prisma app. With advanced models, the outputs look refined, but style transfer is not a new or useful feature. Despite Levoy’s criticism of the prompt-based implementation, Project Indigo has its own feature that lets you describe your edit. Users can use it to perform edit flows that are not available in preset tools. But this is also a potential road to slopland. 有些风格让我想起了 Prisma 应用的早期阶段。得益于先进的模型，输出效果看起来很精致，但风格迁移并不是一项新颖或实用的功能。尽管 Levoy 对基于提示词的实现方式持批评态度，但 Project Indigo 自身也具备允许用户描述编辑需求的功能。用户可以用它执行预设工具中无法实现的编辑流程。但这也有可能导致内容质量的泛滥。

The company is using Google’s Gemini-based Nano Banana for these features, but it’s open to swapping in other models, including its own Adobe Firefly model. These features, bundled under the AI playground tab, are still in a testing phase, and only select users will get access to them. These features might not ever make it to a wider audience, but it is good to see some features that can make people aware of nuances in photography, rather than just creating more slop. 该公司正在为这些功能使用基于谷歌 Gemini 的 Nano Banana 模型，但也持开放态度，未来可能会更换其他模型，包括其自家的 Adobe Firefly 模型。这些功能被整合在“AI 游乐场”（AI playground）标签下，目前仍处于测试阶段，仅部分用户可以访问。这些功能或许永远不会向更广泛的用户群体开放，但很高兴看到一些能让人们意识到摄影细微差别，而不是仅仅制造更多垃圾内容的功能出现。