How to Make the AI Earth Zoom Effect

The AI earth zoom effect is one continuous camera pull-back from a single photo: you, then the place around you, then the city, the coastline and the clouds, until the Earth is visible from orbit. VdoBloom has a dedicated Earth Zoom Out tab that does it in one gesture — upload an outdoor photo and generate, with no cuts anywhere in the clip — or you can build the same move by hand in the Image to Video tab with a pull-back camera prompt. A 4-second 480p clip on Seedance 2 Mini is 4 credits.

Updated October 2026.

What makes the effect work

The reveal is a scale joke, and it only lands because nothing interrupts it. The camera starts close on the person in your photo, standing still, and rises straight up and away — the place around them appears, then the whole city, then the coastline and clouds, until the planet is in frame against black space. The move is smooth and steady rather than dramatic, and the person never moves, turns or changes size the way they started.

Everything in the frame follows from one decision: keep going up, and do not blink. That is why it reads as a journey rather than a sequence of pictures.

Two ways to make it on VdoBloom

The dedicated tab. Earth Zoom Out is a preset effect with the whole move already written into its prompt. There is nothing to compose: upload the photo, set the format, generate. Its default is a longer clip than an ordinary effect, because the reveal has a long way to travel, and it runs on Seedance 2 Fast at 720p in 9:16 by default.
By hand in Image to Video. The Image to Video tab takes any photo and any motion prompt, and its camera-move chips include a pull-back that reads as “slowly pulls back to reveal the wider scene”. Building it manually gives you control over the wording and the model, and covers variations the preset does not.

The same photo also works on the sibling effect tabs — bullet time and dolly zoom.

Step by step with the dedicated tab

Start with an outdoor photo. This matters more here than in any other effect. The pull-back has to reveal a world around the person, so a street, park, rooftop, garden or beach gives it something real to build outward from. An indoor wall gives it nothing.
Stand back so there is room in the frame. A photo where the person fills the frame edge to edge leaves no space for the place around them to exist.
Upload and choose the format. Duration, resolution and aspect ratio are set before you run it. 9:16 suits Reels, Shorts and TikTok; 16:9 suits a YouTube intro or a widescreen edit.
Generate. The camera rises straight up and keeps going. The aerial and satellite views are generated around your starting point.
Watch the middle of the clip at full size. The handover between what is visible in your photo and what is being invented above it is where this effect most often goes wrong. Play it through and check the place stays believable as the camera climbs.
Keep the photo. The same upload runs on the other camera-move tabs, so a set of clips for one image stays visually consistent.

Step by step in Image to Video

Open the tab and upload the photo in the settings panel. The prompt box will accept the motion description while the image is the first frame of the clip.
Write the ascent as one continuous move. Name the person first, then the chain of reveals, then the rule that there are no cuts.
Add a camera chip if you want. Pull-back is the natural one; crane up reads as a mechanical crane rather than a drone.
Repeat the identity instruction. End the prompt by naming the face, hair and outfit as fixed, so the model does not restyle the person while it invents the planet.
Pick the model and length, then generate. Longer clips cost more, and the reveal needs a reasonable amount of time to travel from a portrait to orbit.

The photo matters more than the prompt

The pull-back is a chain of reveals, and every link has to look like the same world as the one below it. Your photo is not just the subject; it is the anchor for a city, a coastline and a continent that will be invented above it. A recognisable setting gives the model something to scale up from; a blank wall forces it to invent everything above the person’s head, which is where generated scale reveals look thin.

Light behaves the same way. The clip keeps the light of the original photo rather than restyling it, so an evenly lit outdoor shot stays believable throughout. A hard flash photo leaves a shadow that rotates as the camera rises and looks wrong by halfway.

Prompt examples

The core move: “One continuous camera pull-back with no cuts: the camera starts close on the person in the photo standing still, then rises straight up and away, revealing the place around them, then the whole city, then the coastline and clouds, until the Earth is seen from orbit as a blue planet against black space. Keep the face, hair and outfit exactly as in the photo.”
Slower, more cinematic: “A single unbroken rising shot from the person upward: street, rooftops, city grid, river, coastline, cloud layer, then the curve of the Earth. Smooth steady motion, no cuts, no camera shake, the person remains still and unchanged in the opening frames.”
Starting from a rooftop: “Begin on the person at a rooftop railing at dusk, then climb continuously past the building, the neighbourhood, the city lights and the coastline until the planet fills the frame. Photoreal aerial and satellite views, one take.”
Beach version: “Continuous pull-back from a person on a beach: the sand, the bay, the coastline, the island, the ocean and finally the Earth from orbit. No cuts, steady rise, warm light preserved from the opening frame.”
As an intro beat: “End wide and quiet so a title can sit over the final seconds: one unbroken pull-back from the person to the planet, smooth and slow, no text or graphics.”

Note what the prompts do not do: they never re-describe the person. The identity comes from the photo.

Common mistakes

Uploading an indoor photo. The most common failure. The reveal needs a world to open out into.
Cropping the person tightly. The camera has nowhere to travel and the result reads as a zoom rather than a journey.
Adding camera drama. Whip pans, shake and fast zooms fight the effect. It is calm by design.
Cutting the clip short. Too few seconds and the planet arrives abruptly. This effect needs longer than a normal clip.
Checking only the first frame. Check the middle, where the transition from your photograph to the generated world happens.
Forgetting who is in the photo. Use your own image, or one from an adult who agreed — the clip opens on that face.

Which model to pick

The earth zoom default is Seedance 2 Fast at 720p. Building it by hand in Image to Video, the choice is mostly cost against how well the face holds in the opening frames:

Seedance 2 Mini — the cheap route. Image-to-video is 4 credits for 4 seconds at 480p, 8 at 720p, and 10 credits for 10 seconds at 480p. See the Seedance 2 Mini model page.
Seedance 2 Fast — the effect tab default; better detail than Mini, and the tier to step up to when the face matters.
Seedance 2 and Seedance 2.5 — 1080p and longer durations; Seedance 2.5 reaches 30 seconds, priced per second by resolution.
Other image-to-video families — Kling, Veo 3.1, Wan 3.0, Hailuo 2.3, PixVerse, Vidu Q3, Runway Gen-4.5, Grok Imagine and P-Video are all on the tab, each with its own durations and prices.

Whichever you pick, the price is quoted before the run and a clip that fails is refunded automatically. More on the choice is in the earth zoom out tool page, and the wider problem of believable motion is covered in making an AI video not look like a slideshow.

What it costs

Clips are priced by model, resolution and length, and shown before you generate. New accounts start with 10 free credits and no card, which covers two 4-second 480p Seedance 2 Mini clips at 4 credits each. The Lite plan is $15 a month for 300 credits, or $10 a month billed yearly; one-time packs start at $10.39 for 200 credits and never expire; India plans start at ₹199 a month. Paid output is watermark-free with commercial rights. See the pricing page.

Frequently asked questions

How do I make the AI earth zoom effect?

Upload one outdoor photo to the Earth Zoom Out tab and generate: the clip starts on the person and rises continuously through the place, the city, the coastline and finally the planet, with no cuts. You can also build it in the Image to Video tab with a pull-back prompt if you want control over the wording and the model.

Does the clip have cuts in it?

No, and that is the point. The reveal is written as one unbroken move, which is what makes it read as a scale gag rather than a series of stock shots. The camera stays smooth and steady from the first frame to the last, and the person stays in the same spot the whole time.

What kind of photo works best?

An outdoor photo of a person in a recognisable place — a street, garden, rooftop or shoreline — shot from far enough back that there is space around them. Indoors against a plain wall is the worst case, because the model must invent the entire world above their head.

How long should the clip be?

Longer than a typical effect, because there is a long way to travel between a portrait and the whole planet. The Earth Zoom Out tab’s default duration is set with that in mind, and you can adjust it along with resolution and aspect ratio before you generate.

Will the person still look like me?

The clip opens on your photo, and the face and clothing come from that upload, so the person stays recognisable at the start. Check the opening frames at full size before using the clip. Only upload photos of yourself or of an adult who has agreed to appear.

Open Image to Video and animate your photo, or use the Earth Zoom Out preset — the credit cost of the clip is shown before it runs.