How to Prompt Sora for Realistic Video (2026)

Realistic Sora 2 video comes down to one thing: writing like a director briefing a crew, not a novelist setting a scene. This guide gives you the exact 5-part formula, the camera and lighting vocabulary Sora 2 actually understands, and 10 full example prompts you can paste straight in. For a shorter reference once you know the formula, keep the Sora prompt cheat sheet open in another tab.

The 5-part shot formula

Every realistic Sora 2 prompt breaks into five beats. Write them in order and keep each one short — this is the single biggest lever for realism.

Stack those five beats into one paragraph of 40–90 words and you have a prompt that reads like a shot list, which is exactly what Sora 2 was trained to follow. Skip any beat and the model fills the gap with generic, floaty motion that reads as synthetic. See the full pack of 35 best Sora prompts for more finished examples built on this same structure.

Camera language that works

Sora 2 responds to real cinematography terms, not vague words like "cool angle." Use exactly one movement per shot — combining moves is the fastest way to get warped, unrealistic motion.

• Dolly in / dolly out — camera moves straight toward or away from the subject
• Tracking / trucking shot — camera moves alongside a moving subject
• Crane / jib — camera rises or lowers on a vertical arc
• Pan — camera pivots left or right from a fixed point
• Tilt — camera pivots up or down from a fixed point
• Push-in / pull-back — a slow, subtle version of dolly in/out
• Handheld — slight natural shake, documentary feel
• Static / locked-off — no camera movement at all, subject moves within frame

Other useful terms: arc, whip pan, steadicam, low angle, high angle, over-the-shoulder, close-up, wide establishing. Pick the one move that matches the emotion of the shot — a push-in for tension, a static locked-off frame for calm observation — and leave it at that.

Lighting & lens language

Light and lens choices are what separate a flat AI clip from something that looks shot on a real camera. Name a specific light source and a lens feel in every prompt.

• Golden hour — warm, low-angle sun, long soft shadows
• Softbox — diffused, even studio light, flattering on skin
• Practical lights — light sources visible in the scene (lamps, neon signs, candles)
• Volumetric light — visible light shafts through fog, dust, or window blinds
• Shallow depth of field — blurred background, sharp subject, like a wide-aperture lens
• 35mm / 85mm — 35mm reads wider and more documentary; 85mm reads tighter and more cinematic/portrait
• Film grain — subtle texture that breaks up the "too clean" digital look
• High-key / low-key — bright and even vs. moody with deep shadows

A good pairing is one light source plus one lens trait plus one texture note, e.g. "golden hour side light, 85mm shallow depth of field, light film grain."

Writing the sound

Sora 2 generates synced audio, so silence in your prompt means silence — or worse, mismatched sound — in the clip. Write sound the same way you write image: specific and short.

• Dialogue — write the exact line in quotes plus a delivery note: she says, "we're almost there," breathless and quiet
• Ambient — describe the room tone or environment: distant traffic hum, rain tapping on glass, café chatter
• SFX — call out specific sound effects tied to the action: the clink of a ceramic cup on the saucer, a door latch clicking shut

Keep audio notes to one or two lines. Overloading a 4–8 second clip with dialogue, ambient, and three sound effects competes for the same short runtime and muddies all of it.

10 example prompts

Each prompt below follows the 5-part formula and ends with a settings note. Copy any of them directly into Sora, or use them as templates and swap the subject.

1. Morning kitchen, handheld

A man in his 30s in a wrinkled t-shirt pours coffee into a chipped white mug. He leans against the counter and takes a slow sip, steam rising. Setting: a small apartment kitchen, early morning light through a half-open blind. Camera: medium shot, handheld, slight natural sway. Lighting: soft morning window light, warm and slightly overexposed, 35mm, subtle film grain. Ambient sound: coffee maker hissing, a spoon clinking against ceramic. (16:9, 1080p, 8s)

Why it works: one subject, one action, one camera move, and ambient sound tied directly to what's on screen.

2. City crosswalk, tracking shot

A courier in a yellow rain jacket jogs across a busy crosswalk carrying a delivery bag. Setting: a downtown intersection at dusk, wet asphalt reflecting traffic lights. Camera: wide tracking shot moving alongside him at street level. Lighting: mixed streetlight and neon signage, volumetric light through light rain, 35mm. Ambient sound: car horns, rain hitting umbrellas, distant sirens. (16:9, 1080p, 8s)

Why it works: the tracking move matches the subject's motion instead of fighting it, and the setting is grounded to a specific time of day.

3. Golden hour portrait, push-in

A woman with windblown hair stands at the edge of a wheat field, eyes closed, facing the sun. Setting: an open field at golden hour, tall grass swaying. Camera: medium close-up, slow push-in. Lighting: warm backlit golden hour sun, lens flare, shallow depth of field, 85mm. Ambient sound: wind through wheat, distant birdsong. (16:9, 1080p, 8s)

Why it works: a single still action (standing, eyes closed) lets the push-in and golden hour light do the emotional work without competing motion.

4. Diner booth, static locked-off

An elderly man in a flannel shirt stirs sugar into a cup of black coffee at a diner booth. Setting: a retro diner interior at night, red vinyl booths, rain streaking the window behind him. Camera: static locked-off medium shot. Lighting: warm practical lights from hanging pendant lamps, low-key, film grain. Dialogue: he says quietly, "some nights you just need a booth and a window." Ambient sound: distant kitchen clatter, rain on glass. (16:9, 1080p, 8s)

Why it works: the static camera keeps focus entirely on the dialogue delivery, which is where the realism needs to land.

5. Mountain trail, drone-style crane

A hiker in a red jacket walks along a narrow ridge trail with a walking pole. Setting: a mountain ridge above the clouds at sunrise. Camera: wide establishing shot, slow crane rising to reveal the valley below. Lighting: cool blue pre-dawn light shifting to warm sunrise gold, high-key. Ambient sound: wind gusts, gravel crunching underfoot. (16:9, 1080p, 12s)

Why it works: the crane move is used for exactly what it's good at — a reveal — rather than stacked with a pan or push-in.

6. Studio product shot, dolly in

A single ceramic espresso cup sits on a dark wooden table, steam curling upward. Setting: a minimal studio backdrop, plain dark surface. Camera: close-up, slow dolly in toward the cup. Lighting: softbox key light from the left, shallow depth of field, high-key, 85mm. Ambient sound: faint steam hiss, quiet room tone. (16:9, 1080p, 4s)

Why it works: a short 4s duration and a single static object give Sora very little room to drift, which is ideal for product-style realism.

7. Rainy vertical scene for Reels

A teenager in a hoodie skateboards down a quiet suburban sidewalk, backpack bouncing. Setting: a tree-lined residential street in late afternoon, light rain just stopping. Camera: tracking shot from a low angle following alongside the board. Lighting: overcast soft daylight, wet pavement reflections, 35mm, light film grain. Ambient sound: skateboard wheels rolling over pavement, distant lawn mower. (9:16, 1080p, 8s)

Why it works: the low tracking angle sells speed and scale, and the aspect ratio is left to the settings note, not the prose.

8. Office window, whip pan intro

A woman in a blazer stands at a floor-to-ceiling office window looking out over the skyline. Setting: a high-rise office at sunset, city lights beginning to turn on. Camera: medium shot, static locked-off. Lighting: warm sunset backlight through glass, silhouette edge, practical desk lamp fill, low-key, 35mm. Ambient sound: muffled city hum through glass, a phone buzzing once on the desk. (16:9, 1080p, 8s)

Why it works: practical and ambient light sources are named separately so the silhouette and the fill light both read as intentional, not accidental.

9. Farmers market, handheld documentary

A vendor in an apron hands a paper bag of peaches to a customer across a wooden stall. Setting: an outdoor farmers market on a bright midday, striped canopy overhead. Camera: medium shot, handheld, subtle documentary shake. Lighting: bright natural daylight filtered through the canopy, high-key, 35mm. Dialogue: the vendor says, "these came in this morning," warm and casual. Ambient sound: market chatter, a distant busker's guitar. (16:9, 1080p, 8s)

Why it works: handheld plus filtered daylight plus overlapping ambient chatter mimics how a real documentary crew would actually shoot this.

10. Night rooftop, tilt reveal

A man in a denim jacket leans on a rooftop railing, city lights spread out below him. Setting: a rooftop at night in a dense downtown skyline. Camera: close-up on his face, slow tilt down to reveal the skyline below. Lighting: cool blue ambient city glow, warm practical string lights nearby, low-key, anamorphic. Ambient sound: faint traffic far below, wind at rooftop height. (16:9, 1080p, 8s)

Why it works: the tilt is used as a single deliberate reveal, and mixing cool ambient light with one warm practical source adds the kind of contrast real rooftop photography has.

For more finished shots built on this same structure, browse 24 Sora prompts for cinematic B-roll.

5 mistakes that break realism

1. Too many actions in one shot. "She walks in, sits down, checks her phone, then looks up and smiles" is four actions for a 4–8 second clip. Pick the one action that matters.
2. Stacked camera moves. "Dolly in while craning up and panning left" confuses Sora 2's motion model and produces warped, floaty camera paths. Use one movement per shot.
3. Vague subject. "A person walking" gives the model nothing to lock onto. Name age, clothing, and one physical detail so the subject stays consistent frame to frame.
4. Prose over direction. Flowery scene-setting ("in a world where the sun kissed the horizon") reads worse than plain shot language. Write like a shot list, not a short story.
5. Expecting prose to set length or resolution. Sora has no in-prompt parameters like Midjourney's --ar. Aspect ratio, duration, and resolution are app/API settings — leave a plain settings note at the end of the prompt instead of trying to write it into the scene description.

Run every prompt through the 5-part formula above before you generate, and cross-check camera and light terms against the Sora prompt cheat sheet if you're unsure which move fits the shot.