Most seedance underwater prompts fail for the same reason land-based prompts rarely do: water is not a backdrop, it is a second physics engine layered on top of the scene. A prompt that says "underwater shot" without specifying how light bends and attenuates with depth, how a body moves against drag and buoyancy instead of gravity, and where the particulates and caustics live in the water column, hands the model an empty tank. What comes back is a diver-shaped silhouette floating in blue haze — technically wet, but with none of the physical weight that makes underwater footage read as real. The prompts that work treat water as a set of named constraints: a light source that has to travel through meters of attenuating medium before it reaches the subject, a body that drifts and corrects rather than walks, and a single, spatially continuous location the camera can move through without breaking the illusion of a real place.
The five prompts below approach underwater video from five different structural angles. One builds an entire evolved species around the constraint of breathing underwater and locks the whole sequence inside one temple for spatial continuity. One uses a slow, unbroken forward dolly as the narrative engine of an underwater documentary. One proves that a single plain sentence, if it names the right genre object, can carry an entire prompt. One turns the ocean into a horror-scale threat viewed through a containment window. And one uses light itself — bioluminescence, shafts, reflections — as the mechanism that reveals a discovery. Together they show that "underwater" is not one register but at least five, each with its own physical vocabulary.
1. The evolved-diver archaeology — single-location continuity as an immersion device
See the full prompt on Shoty →
"Photorealistic underwater sci-fi archaeology film set entirely inside a single colossal submerged temple ruin beneath Veridion."
Why this works: The prompt's most consequential decision comes before any character description: it locks the entire 15-second sequence inside one connected location, the flooded temple sanctuary, rather than cutting between multiple underwater sites. The prompt states this explicitly as a technique, not an accident — "the entire sequence remains in this one connected location for stronger spatial continuity and immersion." Underwater cinematography is unusually vulnerable to spatial confusion because water removes most of the depth cues a viewer relies on above the surface (horizon line, cast shadows, parallax against a stable ground plane), so a location cut underwater disorients far more than a location cut on land. Keeping the camera inside one architecturally continuous space — entry hall, central chamber, relic pedestal — lets the viewer build a mental map even without those cues.
The character design solves a second underwater problem: what does a human diver's breathing apparatus do to the shot? Rather than a scuba tank and regulator (which occupy frame space and break the fantasy of belonging underwater), the prompt gives Nala engineered gills that "rhythmically open and close while breathing underwater," with bioluminescent circuitry that visibly reacts to environmental stimuli. This converts a logistics problem — how does she breathe — into a character beat the camera can hold on in close-up. The gills aren't hidden; they're the reveal.
Takeaway: When an underwater scene needs to sustain more than a few seconds of screen time, name one continuous location and route the entire camera path through it rather than cutting between setups — water's lack of depth cues punishes spatial jumps more than any other environment. If the subject needs to breathe on camera, design the breathing mechanism as a visible character detail instead of hiding it, so it becomes something the camera can use rather than something it has to avoid.
2. The sunken Egyptian city documentary — the locked forward dolly as narrative engine
See the full prompt on Shoty →
"Photoreal cinematic underwater documentary, vast sunken Egyptian city on the ocean floor, hauntingly serene atmosphere, ancient limestone architecture half-buried in sand"
Why this works: This prompt structures its entire 15 seconds around a single camera instruction repeated and refined at every beat: a slow forward dolly that never stops, never cuts, and never pans randomly. "Camera: slow underwater dolly forward, soft stabilized movement, no sudden cuts, no random pans, no shaky motion" is stated as an explicit constraint, and the six timestamped beats are simply what that dolly passes on its way through the city — establishing wide shot, pharaoh statue, sphinx, low pass near the paws, ruined colonnade, temple courtyard. The dolly is not a camera move applied to a scene; it is the scene's organizing structure, converting a static ruin into a documentary reveal with forward momentum.
The negative prompt does more narrative work here than in most prompts of this length: "no modern divers, no submarines, no plastic debris, no fantasy monsters, no glowing magic symbols, no clean new architecture." Each exclusion protects a specific documentary illusion — that this is a real, undisturbed, ancient site rather than a movie set or a fantasy environment. Without banning modern divers and submarines, a documentary-tagged underwater prompt runs a real risk of generating a research vessel or a diver in frame, which instantly breaks the "untouched ruin" premise the whole prompt is built on.
Takeaway: For underwater documentary footage, commit to one camera move for the full duration rather than switching between shot types — a locked, slow forward dolly does the work of narration on its own, turning a sequence of static subjects into a continuous reveal. Use the negative prompt to protect the scene's core illusion (untouched, ancient, undisturbed) by excluding the specific modern or fantastical elements most likely to break it.
3. The vintage brass diving suit — one plain sentence, one unmistakable genre signal
See the full prompt on Shoty →
"A diver in a vintage, heavy brass diving suit with a glass helmet port exploring beautiful scenes under ocean 🌊."
Why this works: Where the temple archaeology and sunken-city prompts run to hundreds of words with timestamped beats, this one is a single sentence — and it still produces a strongly art-directed result, because it front-loads one extremely specific, era-defining object: the vintage brass diving helmet with a glass viewport. That single noun phrase does the work an entire paragraph of lighting and camera direction would otherwise need to do. A brass diving helmet is inseparable from a narrow set of visual defaults — early-20th-century hard-hat diving gear, riveted metal, a small round viewport, tethered air hose, slow deliberate movement because the suit is heavy and pressurized. Naming the object activates that entire visual bundle at once.
This is a useful counter-example to the other four prompts in this set: exhaustive specification is a tool, not a requirement. When the goal is a recognizable historical or genre register rather than a specific narrative beat, one correctly chosen object noun can outperform a paragraph of generic mood adjectives ("moody," "cinematic," "atmospheric") that don't tell the model anything it can render differently.
Takeaway: Before writing a long underwater prompt, ask whether a single well-chosen period or genre object — a brass diving helmet, a rebreather rig, a modern dry suit — would already imply the lighting, pacing, and production design you're trying to describe in separate sentences. Specificity in the subject often buys more than specificity in the adjectives.
4. The abyssal containment breach — scale, and the one-move-per-threat rule
See the full prompt on Shoty →
"A lone male engineer in a reinforced exosuit stands on a narrow steel catwalk inside a deep-sea containment observatory, facing a colossal curved acrylic viewport."
Why this works: This prompt solves the hardest problem in underwater creature horror — how do you convey a threat larger than the frame can hold — by putting the camera and the audience surrogate on the same side of a viewport, and the monsters permanently on the other side. The acrylic window becomes both a physical constraint (something the creatures cannot cross without ending the scene) and a scale reference: everything the viewer judges the leviathan's size against is the engineer standing in front of that same glass. This is a variant of the "known object for scale" technique, except the known object is a piece of architecture rather than a person, which lets the prompt establish enormous scale without ever putting a human body in genuine danger.
The prompt also imposes an explicit choreography rule — "One-Move Rule per titan" — assigning exactly one distinct action to each of the three creatures (ram, claw-drag, electrified wrap) rather than letting all three attack simultaneously in an undifferentiated frenzy. This sequencing is what keeps a three-monster shot legible: the viewer processes the leviathan's impact, then the crustacean's claw, then the jellyfish's electrical wrap, as three discrete beats rather than one chaotic blur. "No glow/bloom VFX" is a deliberate restraint too — it keeps the creature lighting inside the deep-cold-blue, alarm-red practical palette the prompt establishes rather than letting stylized VFX glow break the photorealistic register.
Takeaway: For underwater threats too large to frame directly, put a barrier — a viewport, a cage, a submersible hull — between the camera and the creature, and use a human-scaled object on the camera's side as the scale reference. When staging multiple creatures at once, assign each one a single distinct action rather than letting them all move simultaneously; sequencing separate the threats into readable beats instead of visual noise.
5. The bioluminescent shipwreck discovery — light as the mechanism of reveal
See the full prompt on Shoty →
"A vast underwater abyss stretches endlessly beneath a dark cobalt ocean, faint sunlight cascading down in thin shimmering rays that ripple across jagged seabed cliffs and ancient coral-covered ruins"
Why this works: This prompt layers three distinct light sources across its three acts, and each light source corresponds to a different emotional register. The opening abyss is lit by "faint sunlight cascading down in thin shimmering rays" — a weak, distant, top-down light that establishes scale and isolation by showing how little of the surface world reaches this depth. The middle act shifts the dominant light source to the environment itself: "ethereal turquoise and emerald lighting pulses softly from bioluminescent organisms hidden among the reef," which changes the mood from isolation to wonder because now the space is generating its own light rather than merely receiving scraps of it. The final act introduces a third source — "reflections from faint wreck illumination shimmer across the diver's visor" — which is the payoff light, tying the discovery directly to the diver's face at the moment of contact.
This progression (weak external light → ambient bioluminescent light → close reflected light on the subject) is a compressed three-act structure told entirely through where the light comes from, without a single cut or explicit narrative instruction. The diver's role in the frame also shifts to match: distant and small under the sunlight rays, then integrated into the glowing reef during the bioluminescent middle act, then in intimate close-up for the visor reflection at the shipwreck. Camera distance and light source move together.
Takeaway: Structure a discovery sequence by changing the dominant light source across acts rather than only changing the camera or the subject's action — a shift from distant natural light, to ambient environmental light, to close reflected light on the subject's face mirrors a story arc from isolation to wonder to intimate contact, and it reads clearly even without dialogue or explicit emotional direction.
What these five underwater prompts have in common
- Water removes depth cues, so location continuity matters more than on land. A single connected space that the camera moves through — rather than cuts between — keeps the viewer spatially oriented in an environment with no stable horizon or ground shadows to rely on.
- Name the breathing or life-support mechanism instead of hiding it. Whether it's engineered gills, a brass helmet, or an exosuit, the apparatus that lets the subject exist underwater is a visible character and production-design detail, not a problem to avoid framing.
- Camera movement can carry the entire narrative. A locked, slow forward dolly through a series of fixed subjects functions as documentary narration on its own — no cuts or dialogue required.
- One correctly chosen genre object can replace a paragraph of mood adjectives. A brass diving helmet or a specific creature design activates an entire visual bundle that generic words like "cinematic" or "moody" cannot.
- Scale for creatures too large to frame comes from a barrier plus a known reference object, and multiple simultaneous threats stay legible only when each is given one distinct, sequenced action.
- Light source is a narrative tool underwater, not just an exposure setting — shifting from distant natural light to ambient bioluminescence to close reflected light can carry an entire emotional arc without a single explicit direction.
For adjacent techniques, see the 5 Seedance Nature & Wildlife Prompts for creature behavior and buoyancy physics, and the 5 Seedance VFX & Special Effects Prompts for the containment-breach viewport technique applied to other threat types. The full underwater use-case gallery collects more depth-stratified light and buoyancy prompts, and How to Write Seedance 2 Prompts covers the general prompt-structuring principles behind all five techniques above.
