Skip to content

Tools

Adobe opens AI photo editing previews for painting surfaces and reshaping objects

Edited and reviewed by AI Ping Editorial Team

3-line summary

  1. Adobe opened browser previews for Project Deep Brush and Project Hands On after showing the tools at MAX 2026 Sneaks.
  2. Deep Brush accounts for shape and lighting as users paint. Hands On generates points users can drag to reshape details.
  3. The Scene Sweep video experiment has a notification signup. Adobe gave no date for a full release of the two image tools.
PracticalDesigners
A hand holds a stylus above a pen display showing coral fabric folds
AI-generated conceptual image: A stylus poised over folds of lit fabric illustrates working directly on a photo. · AI Ping

Adobe opened browser previews for two image-editing experiments it presented on November 11 at MAX 2026 Sneaks, the conference’s research showcase. Project Deep Brush lets users paint along the surfaces of objects in a photo. Project Hands On puts controls on objects so users can drag a detail such as a mug handle into a different shape. Both previews are available through Adobe Labs.

The session featured 10 research demos across areas including video and sound. Adobe said this was the first year it had given people a chance to try select projects from the session. The two image projects have links to their own browser editors, both labeled as previews. Adobe gave no date for a full release or for bringing either tool to Photoshop.

Painting a bottle without losing its shadow

Putting a new color or pattern on a bottle in a photograph involves more than replacing the color of its pixels. The same material looks brighter on the side facing the light and darker on the side in shadow. Adobe says Deep Brush estimates the object’s shape, surface and lighting from a single photo and handles that information separately. In a packaging mockup, the idea is that a painted design can follow the bottle’s curve and darken where it passes into shadow.

Adobe describes two kinds of work with the brush. A user can choose where to put new paint or adjust the light falling on the photographed scene. That gives the person editing the mockup a way to treat the pattern and the lighting as separate parts of the job. The research team worked with a professional artist on the controls. Adobe lists garments and walls as other surfaces where the tool could be used to try out a design. Those examples describe the intended workflow. The result depends on what the tool can infer about the scene from the photo.

Adobe had introduced Photoshop’s Relight a day earlier at MAX. Relight lets a user add a light source and adjust its position, color and brightness. In the Deep Brush experiment, the user specifies the area of a photo with a brushstroke. The tool uses its estimates of the surface and lighting to render the stroke.

A turquoise brushstroke continues into the shadow on a curved ceramic bottle
AI-generated conceptual image: A turquoise stroke darkens around a ceramic bottle’s curve, illustrating how paint can follow the shape and shading. · AI Ping

Pulling a mug handle into shape

Hands On separates objects from their surroundings in the image and uses a vision-language model to help place controls on the canvas. Adobe gives the example of a mug handle. Dragging the points can reshape the handle or adjust its proportions. Adobe also sees a use for the controls in product mockups, where a designer might compare small changes in shape before anything is manufactured.

Photoshop already lets people reshape parts of an image by dragging points. In Puppet Warp, the editor places pins on a mesh to tell Photoshop which areas to move and which to anchor. Hands On analyzes the image and proposes the points itself. Adobe says the project grew out of work used in Firefly’s Edit Object tool, where users describe the edit in text. In the new preview, the person editing moves the on-canvas points.

In a separate report, Adobe Labs drew on eight qualitative studies involving 82 established and aspiring creative professionals and reviewed 27 papers. Its researchers described how people use reference images, compare what they see and point to details as they work out what they want. That study examined how people communicate creative intent. It provides no measurements of accuracy or time saved with the two previews.

Two white mugs with differently shaped handles
AI-generated conceptual image: Rounded and angular handles on two mugs illustrate a change to one part of an object. · AI Ping

Which other projects are open to try

Project Scene Sweep, a video experiment from the same session, illustrates the difference in access. It explores how to carry an edit across cuts, such as when the same person appears from another camera angle. Adobe Labs lists a notification signup on the Scene Sweep page, while the two image project pages link to preview editors.

Adobe Labs also lists Project Custom Filters as a separate experiment in its Playground. Adobe is exploring it for Photoshop. A user describes an effect in words, and the experiment builds an adjustable filter that can be applied to the image without changing the original. Adobe gives lens distortion as one example of an effect a user could generate and then refine with the filter controls.

Sources