Adobe Max 大会公布四项 AI 实验工具，照片可走进 3D 场景、AI 角色可对话试镜

Some of the most impressive tech that Adobe showcases at its annual Max conference each year is the experimental features it’s developing, and the latest of those tools are being announced today: they can transform images into explorable 3D scenes, design AI characters by talking to them, and turn photo editing into a copy/paste process.

The first (and weirdest, in my opinion) is Project Screen Test, which is designed to make AI-generated characters easier to develop and use for video content. Instead of just describing how the character should sound and behave in text prompts, the tool lets you talk to the character and direct them in an audition-like environment. You start by describing some base personality traits, such as “anxious, shy, awkward,” and giving the character a voice, either by selecting a preset, describing how it sounds in a prompt, or uploading sample audio for it to reference.

You can then expand its persona in further detail by conversing with the character as if you’re talking on FaceTime, outlining its backstory, temperament, goals, flaws, fears, movement quirks and more. Once you’re happy, the tool lets you generate animated screen tests with the character by describing an environment or scenario you want to put them in.

This is currently a standalone creative exercise, and it’s not clear how (or if) it could feature in any of Adobe’s existing AI video tools. The results provided a more identifiable and consistent personality compared to many soulless animated AI characters I’ve seen, but that’s the entire reason this exists — AI video generation provides far less creative control compared to traditional animation, where every aspect of a character is included with absolute intention.

There’s also Project Sight Span, which reframes photographs and images by turning them into a 3D environment that you can navigate through and reposition. The 3D space generates objects and backgrounds that are outside of the original frame, but feel like a natural part of the image, even if you’re turning to see what’s behind you. While the visuals in the 3D environment are pixelated and ill–defined, the resolution quality is restored when you select your chosen angle.

Sight Span works with any flat image, including photos, paintings, and illustrations, and can preserve and replicate artistic styles. Even repositioned tattoos and facial features in photographs remain surprisingly recognizable and true to the original reference. It’s a bit like a version of Apple’s Spatial Reframing feature on steroids: you can reposition more aggressively and change the shooting point entirely if you want, but it also feels far less deserving of being called a photo afterward.

Project Sound Bender is an AI audio mixing feature that lets you customize sound effects as if you’re clicking your cursor around a color wheel. It works by uploading a video clip and then describing your desired audio in a text prompt, such as “bubbling sci-fi cauldron noises.” Sound Bender then generates sound effects and automatically syncs it to the video, and provides a square sound canvas that lets you adjust the noises in real time. Each corner of the canvas will automatically assign a relevant AI-generated sound, such as dense gurgling with popping bubbles, rumbling boiling rocks, pitched bloops, and crackling electric fire. You can drag your cursor around the canvas to select which noises stand out the most, or customize each corner with different sound effects.

Adobe also showcased Project Rinse Repeat, which lets you copy edits and then automatically apply them to other objects or people in the same photo. For example, if you painstakingly airbrush out acne or blemishes on one person, Rinse Repeat will adapt your specific changes to anyone else in a group of people who need similar touchups. It can also be used to add shadows to products in marketing graphics, even if the objects are different shapes and sizes. Rinse Repeat is built around Photoshop, and the edits you apply to other subjects remain fully editable because the tool recreates them with native Photoshop tools.

These “sneaks” are tech previews of capabilities that might eventually be developed into tools for Adobe’s Creative Cloud apps or Firefly platform, so there’s no guarantee that they’ll graduate into being available features. Almost all of these projects are built around Adobe’s generative AI developments, however, and given AI tools now play a central role in the company’s growth strategy, it’s likely that at least some of the experiments will see an official launch. As wacky or fun as these AI experiments are, they’re also an omen. If they’re useful enough, that means AI may eventually replace the tools and stock libraries that creative professionals currently use, regardless of how complex their feelings are about the technology.

A notable theme across these sneaks is making complicated photo editing, filmmaking, and audio mixing workflows feel more accessible and intuitive, adapting them into something that’s more interactive and easy to use. In some cases the features almost seem like toys, but with the benefit of allowing you to make practical changes to creative projects.

“Making these interactions more intuitive is also about giving experienced creatives more precise control,” senior principle Adobe research scientist Gautham Mysore, told The Verge. “Sound Bender’s sonic exploration capability and Sight Span’s camera positioning are examples of how professionals could apply their judgment and refine a result more directly. That connection between creative intent and control is a big part of what we’re exploring.”

Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.

• Jess Weatherbed