How to Make an Earth Zoom Out AI Video (The Zoom-to-Space Trend, 2026)

What You're Trying to Make

One photo in, and out comes a shot that shouldn't be possible: it opens as an ordinary ultra-wide phone selfie — arm out, 0.5 lens, you standing in the middle of an open plaza on a bright day — and then the camera lets go. It drifts back off your face, keeps rising, tilts down over the rooftops; you shrink to a figure and then to a dot, the coastline appears, cloud slides past the lens, the horizon starts to bend, and the camera breaks through the top of the atmosphere until the whole Earth is hanging in a field of stars. No cuts. One unbroken take, about twelve seconds, with the wind thinning into silence as you climb.

The reason it lands is the first second: it looks like a photo you actually took. The impossible part only starts once the viewer has accepted the ordinary part.

This is the earth zoom out effect, and it's one of the most reliably scroll-stopping formats on the internet right now. This guide covers what it is, why it works, the prompt recipe, and the one-tap way to make yours.

Fastest way — Earth Zoom Out on Starrd bakes the whole pull-back into one tap: upload a photo and it renders the ultra-wide selfie opening, the lift-off, the city, the coastline, the atmosphere break and the final planet — with audio — for 100 credits (about $5.99), no prompt to write and no editor. Want the full method first? Read on. ↓

Is This a Trend? (Yes — and It's Still Climbing)

It's real and it's enormous. #EarthZoomOut has cleared 1.2 billion views across social platforms, and search interest for "AI earth zoom out" runs at roughly 250,000 queries a month worldwide. It broke out through Higgsfield's zoom-out effect — the same tool behind the eye-zoom trend that took over TikTok earlier in the year — and it's since become a stock move in every creator's kit.

Here's why it has lasted longer than most effects: it isn't tied to a joke, a sound or a moment. A dance trend dies when the song dies. A zoom to orbit is a camera move, so it slots into anything — a travel video, a product reveal, a music intro, a "where I'm from" post, a graduation clip. The human eye is hard-wired to follow fast changes in scale, so it grabs attention in the first half-second and then keeps paying it off for twelve.

The catch is that the standard way of making it is genuinely annoying. The common workflow is: screenshot the frame you want to stitch to, feed it into an effect tool, wait, download the clip, drop it in an editor, reverse it, colour-match the exposure to your real footage, and stitch the two together. That's five or six steps and an editor before you have anything postable. A ready-made template collapses all of it into an upload — which matters, because trends reward speed.

The Fastest Way — Use the Earth Zoom Out Template on Starrd

The Earth Zoom Out template is live in the Starrd library. It packages every step in this guide — the ultra-wide selfie opening frame, the continuous ascent, the scale ramp, the atmosphere break and the final planet — into a single upload.

1. Pick one clear photo. Full-body or three-quarter, good light, face visible.
2. Open the Earth Zoom Out template in the Starrd app or web library.
3. Upload the photo and tap generate. It builds the ultra-wide selfie anchor frame from your photo, then renders a 12-second 9:16 vertical pull-back on Seedance 2.0 with synced audio.

One photo, a few minutes, no editor. Want to start somewhere specific? Type it in Director's Notes — a rooftop at golden hour, a beach, a football pitch, your own street, a particular outfit. The zoom-to-space structure holds regardless of where it starts.

Or, Build It Yourself — What You Need

• One clear photo of the subject.
• An image model to build the ultra-wide selfie anchor frame (the first frame of the zoom).
• A video model that holds a continuous camera move across a big scale change — this is the part most models get wrong.
• Optional: an editor, only if you want the reversed zoom-in version or want to stitch it onto real footage.

Step 1 — Pick Your Reference Photo

The earth zoom out is unusual: your face is only legible for about the first second. After that you're a figure, then a dot, then nothing. That makes the opening frame do all the identity work at once, so the photo matters more than in a format where the camera stays on you.

• Full-body or three-quarter beats a tight head-shot — the model needs to place a whole person on the ground from above.
• Bright, even light. The default look is clear midday sun with a short hard shadow directly under you; a photo shot in that kind of light composites far more convincingly.
• One person. Group shots confuse the centring, and the whole effect depends on a single subject dead centre.
• No heavy filters, eyes open, face unobstructed.

Step 2 — Lock the Concept

Before prompting, decide three things:

1. Where it starts. An open space with a strong radial or grid pattern is ideal — a circular plaza, a running track, a crossroads, a rooftop, a wide beach. Patterns that radiate outward from the subject give the pull-back something to read scale against, which is what sells the motion.
2. The camera angle. Start at eye level on an ultra-wide 0.5x lens, held slightly above head height and angled down, with the subject looking up into it. This is the counter-intuitive bit: an overhead drone frame seems like the obvious opener, but it makes the subject look composited — dropped into the scene rather than standing in it. An ordinary wide selfie reads as a real photograph, and the camera can rise out of it. Let the tilt to top-down happen during the climb, not before it.
3. Where it ends. Full planet against stars is the classic. You can also stop at the curve of the atmosphere for a subtler, more cinematic version.

Step 3 — Write the Prompt

An earth zoom out lives on one continuous camera move and a scale ramp that escalates every few seconds. Build a clean, believable wide-selfie anchor frame first, then let the video ascend out of it. Copy this and swap in your specifics:

ANCHOR FRAME (single still): an authentic ultra-wide phone selfie, vertical 9:16, with the exaggerated perspective of a 0.5x ultra-wide lens held slightly above head height and angled down. The subject (from the uploaded photo) stands alone in the middle of a wide open city plaza on a clear bright day, looking up into the lens with a relaxed natural expression. Strong ultra-wide barrel distortion — head and shoulders large and close in the upper-centre of the frame, body tapering away below, while the patterned plaza stonework, benches, small trees, surrounding streets and low rooftops sweep outward and downward around them, and a big open sky fills the top of the frame. Bright midday sun, a touch of lens flare, the casual imperfect look of a real phone photo.VIDEO (12s, vertical 9:16, ONE continuous unbroken take — begins as a handheld ultra-wide phone selfie and becomes a satellite zoom):[0–3s] Ultra-wide selfie framing — the subject looks up into the lens with a small smile and a little wave, hair stirring in the breeze. The camera begins to drift smoothly backwards and upwards away from their face; the tight selfie framing opens out and more of the plaza comes into view. Continuous smooth pull back.[3–6s] The camera keeps rising and tilts down into a top-down view as the subject shrinks beneath it to a small figure and then to a dot. Streets, rooftops, parks and moving traffic spread out below until the whole city fills the frame. Continuous ascent, no cut.[6–9s] The ascent accelerates hard. The city collapses into a grey sprawl; coastline, rivers and mountain ranges resolve below. Wisps of white cloud slide past the lens and the horizon begins to curve. Continuous vertical ascent, no cut.[9–12s] The camera breaks through the top of the atmosphere into black space. The curved blue limb of Earth swings into frame, swirled with white cloud and lit hard from one side, then settles as the whole planet hanging in a field of stars. Motion eases into a slow drift.Audio: soft plaza ambience and breeze at the start, a rising cinematic swell and deep sub-bass as the camera climbs, thinning wind and air, then near-silence with a low space drone at the end.Avoid: cuts, scene changes, returning to the subject, the subject drifting off centre, camera shake, morphing, distortion, sudden scene jumps.

The non-negotiable elements:

• An ordinary-looking opening frame. A real-feeling wide selfie, not a drone shot. It buys the viewer's belief before you spend it.
• The tilt happens during the climb. Eye level to top-down as the camera rises, described as one move — never as a new shot.
• Escalation every segment. Person, block, city, region, continent, planet. Asking for "zoom out to space" in one leap gets you a cut or a slideshow.
• The subject locked dead centre. If they drift, the ascent reads as a pan and the sense of leaving the ground evaporates.
• An accelerating pull-back. Constant-speed zoom looks flat. The scale is exponential, so the motion should be too.
• Audio that thins out. Ambience to wind to near-silence is what sells altitude more than any visual.

Step 4 — Pick a Model

Seedance 2.0 is the pick. The hard requirement here isn't visual quality, it's camera discipline — holding a single unbroken move through a 10,000× scale change without cutting away — and Seedance has the strongest prompt adherence for exactly that. It also generates the wind, the swell and the silence as synced audio in the same pass, which the format leans on heavily.

Other video models can render the individual looks beautifully, but they're much more likely to cut to a generic stock shot of Earth somewhere around the eight-second mark, and you'll burn renders fighting it.

Step 5 — Generate and Iterate (Common Failure Modes)

• It cut to a stock Earth shot. The most common failure by far. Strengthen the continuity language, put "cuts" and "scene changes" in the negative constraints, and make the middle segments more granular so there's no gap big enough to cut across.
• The subject slid off centre. Add "subject locked dead centre" to the first segment and "the subject drifting off centre" to the avoid list.
• It zoomed but never left the atmosphere. The last segment isn't explicit enough. Name the atmosphere break, the curved blue limb, and the star field as separate beats.
• The face was mush in frame one. That's the photo, not the prompt. Go back to a brighter, sharper, front-facing shot.
• The motion strobed. Usually too much shake stacked on fast motion. Keep the ascent smooth and drop any handheld language — this is a locked, mechanical camera move.

Step 6 — Post It

The earth zoom out is a hook or a closer, rarely the whole video. Two placements that work:

• As a cold open, reversed — the camera falls out of orbit and lands on you, then cuts to your actual content. Two seconds of that buys you the whole first act.
• As a closer, played forward — you say your line, the camera leaves the planet, end card.

Add a trending sound over the top, label it as AI-generated, and don't over-think the first frame — the fact that it looks like a completely normal selfie is the point. The viewer scrolls past a photo, then the photo starts leaving the planet.

Common Mistakes That Tank Your Video

• Opening on a drone shot. The most common instinct and the one that quietly kills it: a top-down frame makes the subject look pasted into the scene instead of standing in it. Open on something that looks like a real photo and let the camera earn the altitude.
• Asking for the whole zoom in one instruction. Without timed, escalating segments you get a jump cut or a slideshow of three scales.
• A busy starting location. Crowds and clutter make the subject impossible to track as they shrink. Empty and patterned beats interesting and full.
• Forgetting the audio arc. A zoom to orbit with constant-volume ambience feels like a zoom on a map. The thinning air is half the effect.
• Over-editing the colour. If you're stitching to real footage, match the AI clip to your footage, not the other way around.

Frequently Asked Questions

What is the earth zoom out AI trend? It's the viral effect where a single continuous shot starts tight on a person and pulls back without ever cutting — off the ground, past the rooftops, through the clouds, out past the curve of the atmosphere, until the whole planet hangs in black space. The appeal is the scale shift: a face to a dot to a pale blue marble in about twelve seconds. The hashtag has cleared well over a billion views, and it slots into almost any video as a hook, transition or closer.

How do I make one of myself? Either the manual route — generate a zoom-out from a still, bring it into an editor, reverse and colour-match it, stitch it to your footage — or the Earth Zoom Out template, which renders the whole continuous pull-back with audio from one uploaded photo, no editor involved.

Why does mine cut instead of staying continuous? Because the prompt didn't forbid it. Models treat a big scale change as a natural cut point. State "ONE continuous unbroken take" up front, list "cuts" and "scene changes" as things to avoid, and escalate scale gradually across timed segments.

Zoom out or zoom in? Same render — reverse the clip in your editor for the arrival-from-space version. Many creators use the zoom-in as a cold open and the zoom-out as the closer.

What photo works best? Clear, bright, full-body or three-quarter, face visible, one person, no heavy filters. Your face is only readable in the first second, so the opening frame carries all the identity.

Which model should I use? Seedance 2.0 — it holds the continuous camera move across the scale change, which is the genuinely hard part, and generates the audio arc in the same pass.

Can I change the starting location? Yes. Director's Notes can move it to a rooftop, beach, field, stadium or your own street, and change the time of day or your outfit, without breaking the zoom structure.

Do I have to label it as AI? Yes — TikTok, Instagram and YouTube all require it, and it costs the video nothing, since a zoom to orbit is obviously impossible anyway.

Related Reading

• AI Camera Movement Prompts — the full vocabulary for directing a camera in an AI video, including pull-backs, drone moves and dolly zooms.
• Viral AI Video Trends in 2026 — the running roundup of what's actually working right now.
• Seedance 2.0 Prompt Guide — how to structure timed segments, negative constraints and audio direction.
• How to Make a Skyfall AI Video — the other big vertical-motion self-insert trend.
• How to Make a Red Carpet AI Video — a multi-beat self-insert with synced voiceover.