Adobe在其年度Max大会上发布了Photoshop、Premiere Pro和Lightroom的最新生成式AI工具。Photoshop的生成式填充功能现可集成Google和Black Forest Labs的第三方AI模型，为用户提供更丰富的图像编辑选项，并能根据文本描述插入、移除或修改内容。此外，Photoshop网页版将推出AI助手（私密Beta），允许用户通过自然语言指令进行图像编辑。Lightroom则引入“辅助筛选”功能，帮助摄影师快速从中筛选出最佳照片。Firefly AI图像模型也得到升级，支持更高分辨率生成，并优化了对真实人物的渲染能力，同时为基于提示词的编辑和图层编辑工具提供支持。

🎨 **多模型支持的生成式填充**：Photoshop的生成式填充功能现已支持集成Google的Gemini 2.5 Flash和Black Forest Labs的Flux.1 Kontext等第三方AI模型，与Adobe自家的Firefly模型一同提供，用户在进行内容添加、移除或修改时，可以根据提示词选择不同模型生成的多种结果，极大地扩展了创作的可能性和多样性。

🤖 **Photoshop网页版AI助手（Beta）**：该功能以类似聊天机器人的界面形式出现，用户可以通过简单的文本指令，如“提高饱和度”，来自动执行复杂的图像编辑操作，Photoshop能够智能运用其丰富的工具集来完成用户需求，从而简化编辑流程，提高效率。

📸 **Lightroom辅助筛选功能**：这项新功能旨在帮助摄影师处理大量照片。它能够根据焦距、角度和清晰度等因素自动筛选照片集，并推荐最适合进行后期编辑的优秀照片，显著减轻摄影师在海量素材中寻找最佳照片的负担。

✨ **Firefly AI模型升级与新功能**：Firefly Image 5模型现可原生生成四百万像素分辨率的图像，无需后期放大，并且在渲染逼真人物方面得到了显著优化。该模型还驱动着新的基于提示词的编辑功能，以及Photoshop中即将推出的图层编辑工具，后者能进行精确、上下文相关的调整，例如在移动对象时自动修正阴影。

Adobe has kicked off its annual Max event, giving us a first look at new and upcoming generative AI tools launching for the company’s Photoshop, Premiere Pro, and Lightroom Creative Cloud apps. These include updates to Photoshop’s Generative Fill feature that aim to give creators more control over adding, removing or modifying content, and tools that can automate some of the more time-consuming elements of editing photos and videos.

To start, Adobe is allowing Photoshop users to power Generative Fill capabilities using Google and Black Forest Labs’ third-party AI models. After selecting their image and giving Generative Fill a prompt — such as describing an object to insert, or replacing an existing object or person with something else — users can switch between Google’s Gemini 2.5 Flash, Black Forest’s Flux.1 Kontext, and Adobe’s Firefly image model, providing a wider variety of results to choose from.

Photoshop for web is also introducing an AI Assistant in private beta. This is a chatbot-like interface that can be given descriptive instructions like “increase the saturation” to automatically edit files, effectively using Photoshop’s vast selection of tools for you. This capability was teased in April, and a similar feature is launching for Adobe Express in public beta.

Another beta feature called “Assisted Culling” is launching for Adobe Lightroom that can filter large collections of photographs for different focus, angles and degrees of sharpness, and recommends the best snaps for photographers to edit.

The Firefly AI image model that powers these editing tools is getting a model upgrade. Adobe says that Firefly Image 5 can generate images in a native four-megapixel resolution without upscaling, and has been optimized to improve its ability to render realistic humans. The model also powers Adobe’s new prompt-based editing features that make specific adjustments based on descriptions, and a new Layered Image Editing tool coming to Photoshop that makes precise, contextual changes, such as ensuring shadows are automatically corrected when moving objects around an image.

For video editors, Adobe also has a new tool for its Premiere Pro software that automatically stencils out people and objects in video frames. AI Object Mask is available in public beta, and aims to make it easier to quickly color grade, blur, and add visual effects to moving backgrounds by removing the need to manually mask subjects using the pen tool.