Adobe Project Sight Span: step inside a photo and reframe the shot
Project Sight Span, shown at the Adobe MAX 2026 Sneaks, turns a single image into a navigable 3D scene you can reframe as new stills or camera moves. How it works, videos, an infographic and when it might reach Photoshop.
Project Sight Span is an Adobe Research experiment, shown at the Adobe MAX Sneaks on 11 November 2026, that turns a single image into a navigable 3D scene: you move the camera, step past the edges of the original frame and generate a new still or a video with a controlled camera move.
- What it is
- Research prototype (Adobe Sneak), not a shipping feature
- Shown at
- Adobe MAX 2026, Miami Beach, Sneaks session on 11 November
- Input
- One image: a photo, painting or illustration
- Output
- A new high-resolution framing or a camera-controlled video
- Team
- Yiwei Hu (lead), Zhiqin Chen, Ziwen Chen, Élie Michel, Skylar Wurster, Kalyan Sunkavalli
- Availability
- No release date, not in beta
Imagine having a single photo, or a painting, and being able to step inside it: move sideways, raise the camera, turn towards what the photographer never framed. That is what Project Sight Span shows, one of the ten experiments Adobe presented on the Sneaks stage at Adobe MAX 2026. In the clips going around social media it is often called “the new Photoshop feature”, but let’s be clear from the start: for now it is a research prototype, and Adobe has not said when, or whether, it will reach its apps.
In recent years, Adobe has shown a strong ability to spot and follow the technology trends of the moment. Today that trend has a very specific name: artificial intelligence. Much of the company’s investment, development and communication now seems to revolve around generative AI.
Personally, I am not against artificial intelligence as a tool that supports creative work. On the contrary, I think it can open up interesting possibilities and expand the toolset available to photographers, designers and artists. The problem, in my view, lies elsewhere: in chasing this transformation, Adobe risks gradually losing sight of the needs of its own community.
People who have been using Photoshop and the other Adobe tools for decades are not necessarily looking to be replaced by a technology; they want to still feel at the centre of the creative process. They want better tools, more control and, above all, to be recognised as professionals and artists, not simply as users to be constantly offered new AI-based features.
It is a matter of trust and identity. Behind Adobe’s products there is a community that helped build the company’s success over the years and that today, in some cases, seems to feel somewhat sidelined. That said, we will see how the community responds to all this technology.

What is Adobe Project Sight Span?
Sight Span starts from one image and turns it into an interactive environment you can navigate. No multi-angle scans, no source video: one frame is enough. The system rebuilds a lightweight 3D scene, good enough to explore framings within seconds (Adobe mentions roughly 10-20 seconds of generation), and then uses a separate image-generation step to render the chosen view at final quality.
The most striking part is that the camera is not trapped by the frame: you can move past the edges of the original photo and the system generates objects and backgrounds that were not in the image, including what would sit behind the camera.
How it works, step by step
The infographic below sums up the process using Adobe’s own demo image, Vermeer’s The Milkmaid:

- Load a single image. It can be a photograph, a painting or an illustration.
- A lightweight 3D scene is created. Think of it as a proxy of the environment, built for fast navigation rather than good looks.
- Move the virtual camera. Orbit, dolly, change height and point of view. Areas outside the original frame are generated on the fly.
- Soft preview, sharp render. While you navigate the image looks soft; once you pick the angle, a generative model recreates it at high resolution from the original image and the camera position.
- Export. The result can be a new still image or a video with a controlled camera move.
In the demo, faces and recognisable details such as tattoos stay consistent in the new angles, and a painting’s style carries over too.
Watch the presentation
Sight Span was shown during the Adobe MAX Sneaks session, hosted this year by actor Terry Crews. Here is the full official recording, with all ten experiments:
Two short reels did most of the rounds on social media. The first is by Justin Odisho, who describes it as a way to step inside a photo, move the camera around and then upscale the result:
The second comes from Indian tech outlet Digit, with a quick recap:
What could it be used for?
Adobe lists several possible uses, very close to the work of 3D, motion and film artists:
- Set design and world building: start from a concept and expand the world around it.
- Storyboarding and previs: try different shots from a single reference image.
- Location scouting and virtual production: preview viewpoints of a place or create backdrops.
- Products: change the angle of a product still without reshooting it.
- Concept art: explore composition variations from the same painting.

Is it the same as Gaussian splats?
Not quite. Creative Bloq compares it to Gaussian splat techniques, which can already rebuild navigable scenes, but usually from many photos or a video. Sight Span works from a single frame and has to invent the geometry and details it cannot see. Other commentators have compared it to Apple’s spatial reframing of photos.
That is exactly why, as DPReview points out, the new angles are generated interpretations, not new photographs of the scene: great for exploring and communicating an idea, less suitable when you need to document what a place really looked like.
When is it coming to Photoshop?
There is no date. Sneaks are Adobe Labs research projects: over the years some have become real features (the official Sight Span page lists related tools such as 360° images in Firefly, VR editing in Premiere and Rotate objects in Photoshop), while others never left the lab. Right now Sight Span is not even in the Adobe Labs Playground, where two other Sneaks from this year, Deep Brush and Hands On, can already be tried.
The other Adobe MAX 2026 Sneaks
Sight Span was in good company. The other nine experiments were Scene Sweep, Screen Test, Live Set, On Cue, Sound Bender, Deep Brush, Hands On, Rinse Repeat and Link Up, across video, audio and design. The common thread, as The Next Web notes, is moving away from text prompts and back to brushes, drags and gestures.
If you are into tools that build scenes from images, also read our article on Blender 5.3 and its Gaussian splat support.
Sources
Official Project Sight Span page on Adobe Labs; Adobe MAX 2026 Sneaks session; DPReview; Creative Bloq; The Next Web; Digit. Videos: Adobe, Justin Odisho, Digit.