The social media trend
A zoom out revealing that the scene was a screen, a poster or a reflection: the effect that makes people watch the video a second time.
Make an infinite zoom video with AI
Your photo becomes the screen of a phone in another setting, then a continuous zoom out reveals the image within the image. The workflow is shown here, no account needed.
Real workflow: Frank at a café terrace in Lisbon, the same photo on the screen of a phone held in a bookshop at night, then five seconds of zoom out from one image to the other.
At a glance
- You provide
- 1 photo · a text or a brief
- What the workflow produces
- 1 image in 4:3 · 1 video of 5s in 720p, 4:3
- Canvas structure
- 6 nodes wired by 5 connections
- Models wired in
- Nano Banana Pro, Wan 3.0
- Cost of one full run
- about 79 credits, roughly $0.95
- Account
- The demo is visible without an account; generating requires a free account, trial credits included
The infinite zoom fascinates on social media: you think you are watching a scene, the camera pulls back, and the scene was only an image within another, then another. This workflow makes the first level from a single photo: it becomes the screen of a phone held in another setting, and the video pulls back from one to the other.
The hard part is the continuity of the pull back. If the phone screen does not show exactly the photo, the model makes a crossfade instead of a zoom. The end frame therefore reuses the photo identically, and Wan 3.0 gets the photo as start frame and the screen as end frame.
It is the same mechanic as the Vertigo effect; for a zoom from very high up, see the zoom from space. The image and the zoom cost seventy-nine credits.
Actual renders from the demo
These are the files of the workflow above, exactly as the models produced them, unretouched. The starting point is shown when there is one.
Starting point
Render
Who it is for
Content creators, music video makers, photographers and app brands who want a reveal effect.
Which files to provide
A strong photo, and the new frame that will contain it.
How it works
- 1
Pick a strong photo
It is what people see first: a face, a landscape, a striking scene, full frame.
- 2
Put it inside another frame
Phone screen, photo on a wall, painting in a museum, reflection in a window: the end frame shows it identically, smaller.
- 3
Give the start and the end
The photo as start frame, the new frame as end frame. Wan 3.0 pulls back from one to the other.
- 4
Chain the levels
Start again from the end frame for a new pull back, then edit the videos end to end: that is the infinite zoom.
Variations and use cases
The same canvas, tuned for different needs: each variation is two or three lines of prompt away.
The music video
Several levels chained to the rhythm of a song, each pull back timed on a strong beat, for an opening or a chorus.
The app ad
The photo of a piece of content becomes the screen of the user's phone: the product reveals itself in the very gesture of the pull back.
The photographer's portfolio
A photo from the series becomes a print hanging on the wall of an exhibition, then the video pulls back to the whole room.
The opening of a video
The first five seconds of a YouTube video: the scene, then the pull back that reveals where we really are, to hold the viewer from the start.
The travel memory
The holiday photo becomes a postcard on a table, then the table a café, then the whole city, level after level.
Common mistakes
Letting the screen change the photo
If the end frame shows another photo, the model crossfades. Ask for the photo identically, twice.
Forgetting the end frame
Without it, Wan 3.0 pulls back towards an invented setting, and the reveal falls flat.
Making the screen too small
A tiny phone in the end frame makes the pull back too fast. A third of the width gives a readable zoom.
Mixing two movements
A pull back and a rotation at the same time blur the effect. One movement, in a straight line.
Recommended models
Nano Banana Pro
Wan 3.0
Kling V3
Seedance 2.5Frequently asked questions
Can I zoom forward instead?
Yes: swap the two images. The camera then dives into the screen until the photo fills the frame.
How many levels can I chain?
As many as you want, one Image node and one Video node per level, edited one after the other in a Montage node.
Does the photo stay sharp on the screen?
In the demo, yes: the screen shows the same terrace. Check the end frame before running the video.
How much does the demo cost?
Seventy-nine credits: nineteen for the end frame on Nano Banana Pro, sixty for five seconds of Wan 3.0 in 720p.
Why an end frame?
It imposes the pull back: Wan 3.0 only goes from the photo to the screen that contains it, without inventing.
Is the zoom continuous?
In the demo, yes: one single pull back, no cut and no crossfade.
Can I keep sound?
Yes: turn sound on and describe the atmosphere of each place. The price does not change on Wan 3.0.
What about a longer zoom?
Set the Video node to ten seconds: the pull back slows down accordingly.
Make an infinite zoom video with AI
Use this workflowFree account, no card required. The workflow opens pre-filled in your canvas.