Sci-fi and cyberpunk video prompts fail in a very specific way: they describe a world but forget to describe a person in it. You get neon streets, flying cars, holographic billboards — and a figure that could be swapped out for any other figure without the shot changing. The character reads as a costume, not a person, because the prompt never specifies what makes them this particular augmented human rather than a generic sci-fi placeholder.
The prompts that work do the opposite. They treat the character like a design document: what augmentations are present and how do they manifest physiologically, what is the emotional target before any action starts, what does the character's silhouette look like against the environment, and — critically — what holds constant when the environment transforms around them. The five prompts below represent five distinct strategies for answering those questions.
Here are 5 Seedance sci-fi and cyberpunk character prompts from the Scenic gallery — from a 25-like post-apocalyptic superhuman fight to a 22-like motorcycle-to-dragon transformation and a neon-street gunslinger pure character study. All free to copy.
1. The augmented superhuman — physiological state as character definition
See the full prompt on scenic.sh →
"Radiation-enhanced superhuman speed and strength, awakened orange-yellow glowing pupils, veins bulging. intense aura. Scene: Abandoned Chicago high-rise ruins, exposed rebar, cracked concrete, suspended derelict vehicles, rusted steel walkways."
Why this works: At 25 likes — the highest in this set — this prompt defines its sci-fi character through biological evidence rather than tech vocabulary. "Awakened orange-yellow glowing pupils, veins bulging" are physiological symptoms of the augmentation, not costume details. The character's powers are visible in their body before any action starts. "Radiation-enhanced superhuman speed and strength" names the power type, but it's the glowing pupils and swollen veins that make it legible on camera in a single frame. The shot sequence then uses that established state to anchor a technically demanding action chain: "Side-dashes at extreme speed with strong motion blur afterimage / kicks off suspended vehicle in mid-air, rotates at high speed / three rapid heavy punches to monster's chest and jaw, each blow forcing it back." Each move is described with a camera instruction attached — "Ground-level chase track shot, full motion blur / Hitchcock spiral orbit shot, high-altitude danger / Medium close-up tracking, 120fps slow-motion" — so Seedance knows both what is happening and how to frame it, not just one. "Desaturated cold gray palette with orange-yellow accent" locks the augmentation color to the visual palette: the character's power signature is also the film's accent color.
The takeaway: define a sci-fi character's augmentation through physiological symptoms (glowing pupils, vein prominence, altered skin) not technology labels. Pair each action beat with an explicit camera instruction so the model knows both what to show and how to frame it. Lock the augmentation's color to the film's accent palette — if the power glows orange, the only saturated element in a desaturated frame should be orange.
2. The transformation arc — identity stability through environmental upheaval
See the full prompt on scenic.sh →
"Feature a female lead character with a strong, distinctive face and unwavering identity across all shots — sharp eyes, confident presence, no variation in appearance. Use a single futuristic motorcycle design — sleek white body, intricate gold mechanical detailing, radiant energy core."
Why this works: At 22 likes, this prompt's core challenge is the hardest in AI video: keep one character recognizable across an environment that transforms completely. Motorcycle rider → dragon rider is not just a visual escalation — it is a continuity test that requires Seedance to treat the character's face and silhouette as fixed points while the surrounding object morphs. "Unwavering identity across all shots — sharp eyes, confident presence, no variation in appearance" states the constraint explicitly, giving the model a hierarchy: character identity supersedes cinematic variety. The bike design specification is equally strict: "sleek white body, intricate gold mechanical detailing, radiant energy core, subtle dragon-inspired contours" are the constants that must carry through the transformation, so the dragon's wings should inherit the gold detailing and the energy core should become the dragon's glowing chest. "While still in motion, the motorcycle begins a controlled transformation. Panels shift, mechanical components unfold smoothly. The front elongates" describes the transformation as a mechanical engineering process, not a magic trick — each component has a logical origin point and a logical destination shape. The closing instruction "camera pulls away to reveal the vast neon city below" frames the reveal as a scale moment rather than an action moment: the dragon and rider are a single composed figure against the megacity.
The takeaway: for a transformation arc, establish two fixed constants before the transformation starts — the character's face and the object's design signature (color, material, light source). The transformation works if those constants persist into the new form. Describe the transformation mechanically: "panels shift, components unfold, front elongates" are engineering verbs that tell Seedance the physics of how the change happens, not just that it happens. Frame the completion of the transformation as a scale reveal, not an action beat.
3. The character-vs-creature structure — antagonist design as character foil
See the full prompt on scenic.sh →
"CHARACTERS: Armored courier, cracked respirator, low baton. Eight-legged mechanical-organic revenant with fractured porcelain face, cable hair, chrome limbs, wet sinew, red pupils, spider-head inside split shell. EMOTIONAL TARGET: Hush into pursuit terror, then capture."
Why this works: At 20 likes, this prompt's structural move is to treat character and antagonist as a designed contrast pair. "Armored courier, cracked respirator, low baton" is a three-noun description — minimal, functional, worn. The baton is low (defensive, not aggressive). The "eight-legged mechanical-organic revenant with fractured porcelain face, cable hair, chrome limbs, wet sinew" is its opposite in every dimension: eight limbs vs two, porcelain (fragile material) vs armor (protective), organic sinew vs mechanical chrome. The antagonist is also an assemblage — spider-head inside a split shell — not a unified form. This design contrast encodes the power dynamic without stating it directly. "EMOTIONAL TARGET: Hush into pursuit terror, then capture" is a separate instruction from the action sequence — it is the emotional through-line the model should pursue underneath the technical actions. The venue is "shuttered pachinko hall, mirrored ceiling, ankle-deep water, emerald spill" — the mirrored ceiling doubles every action and gives the revenant a second axis to move on (ceiling vs floor), which the timeline exploits: "revenant streaks across the reflection... then vanishes on the turn / climbs back up, skitters onto ceiling seams ahead of the courier / drops from the ceiling seams into his path." The environment is not background — it is an active tool for the creature.
The takeaway: design your antagonist as the structural opposite of your protagonist — where the hero is minimal, the antagonist is compound; where the hero is armored, the antagonist is organic-mechanical. State an "EMOTIONAL TARGET" as a separate field from the action sequence — it tells Seedance the arc the scene should pursue emotionally, not just visually. Choose an environment with a geometry the antagonist can exploit (mirrored ceiling, flooded floor) that a human character could not.
4. The 90s anime cyberpunk chase — style specification as character direction
See the full prompt on scenic.sh →
"Rogue courier using @image1 as the exact face reference (identity must remain consistent, adapted into 90s cyberpunk anime style, no facial distortion or drift). Style: Gritty 90s cyberpunk anime noir. High contrast, heavy rain atmosphere, cinematic action framing."
Why this works: At 12 likes, this prompt demonstrates that style specification is a form of character direction. "Adapted into 90s cyberpunk anime style, no facial distortion or drift" gives the model a clear stylization target while asserting that stylization should not compromise identity. The character becomes a specific type of person — a 90s anime courier — not just a person in an anime setting. The lighting structure is written as a three-source system per scene: "[0–3s] Cyan neon rim light outlining the silhouette / [3–6s] Strong red volumetric beams cutting through rain / [6–9s] Deep shadows interrupted by a moving red highlight / [9–12s] Intense white flash from the laser discharge / [12–15s] Fiery orange burst illuminating heavy steam." Each shot has exactly one dominant light source; sources don't compete. The action structure is a classic cyberpunk grammar: hides → spotted → evades → near miss → concealment. Each beat transitions cleanly into the next. "Cybernetic arm twitches and emits unstable blue electrical sparks" at the opening establishes the character's condition before any external threat arrives — the courier is not fine; the arm is already unstable. That detail creates internal stakes separate from the drone threat.
The takeaway: treat a style direction (90s anime noir) as a character direction — it tells Seedance not just how the image should look but what kind of person the character is. Write lighting as a single dominant source per shot (cyan rim / red volumetric / white flash); never describe two light sources with equal priority in the same shot. Establish the character's internal condition (unstable augmentation, pre-existing injury) before the external threat appears — it creates stakes the environment can't produce.
5. The neon-street gunslinger — micro-detail characterization
See the full prompt on scenic.sh →
"2–4s: Slow push-in — subtle mechanical details visible: one glowing eye, faint light lines under the skin, small energy pulses running through cybernetic arm. 6–8s: VFX moment — the eye briefly intensifies with a scanning glow, a thin holographic flicker passes across vision, reflecting in the rain."
Why this works: At 10 likes, this prompt solves a specific problem: how to make a single character standing still in a neon street cinematic for 15 seconds. The answer is micro-escalation — a series of small reveals, each one adding a layer of detail to the same character in the same position. The opening wide shot establishes form (lone figure, long coat, neon signs). The push-in reveals technology (glowing eye, light lines under skin, energy pulses). The VFX moment reveals capability (the eye scans, the hologram flickers in the rain). The draw (8–10s) reveals threat. The shot (10–12s) reveals power. The aftermath (13.5–15s) reveals composure. Six beats, six reveals — all from the same character in the same location, without a single scene change. "Restrained VFX (glow, scan, energy ripple), grounded motion, no text, no overlays, stable proportions without stretch" is the critical constraint: this is a photorealistic study, not a VFX reel. Each tech element is physically grounded (the scan reflects in rain; the energy ripple is "subtle distortion wave"). "Camera: slow cinematic push-ins, minimal cuts, controlled framing, slight handheld micro movement" makes the camera the escalation mechanism — it moves closer as the reveal progresses, so discovery and proximity are synchronized.
The takeaway: build a 15-second sci-fi character study as a micro-escalation sequence — each beat reveals one additional layer of the character's technology, capability, or composure. You do not need multiple locations or actions; you need a discovery structure. Use camera proximity as the escalation axis: start wide (form), push in (technology), hold (capability reveal), pull back (aftermath composure). Keep every VFX grounded in physics — "reflects in rain," "distortion wave expanding forward" — so the technology reads as real, not as a rendering artifact.
Sci-fi cyberpunk character prompt cheat sheet
Across all five, the structural techniques that make Seedance sci-fi and cyberpunk character prompts work:
- Define augmentation through physiology, not technology labels — "orange-yellow glowing pupils, veins bulging" is visible on camera; "radiation-enhanced superhuman" is not. Pair each action beat with a camera instruction. Lock the augmentation color to the film's accent palette.
- Establish two fixed constants before a transformation — the character's face identity and the object's design signature (color, material, light source). Describe transformation as an engineering process using mechanical verbs. Frame completion as a scale reveal, not an action beat.
- Design antagonists as structural opposites — where the hero is minimal, the antagonist is compound; where the hero is protected, the antagonist is organic-mechanical-hybrid. Set an "EMOTIONAL TARGET" as a separate field. Use the environment's geometry as a tool the antagonist can exploit.
- Use style direction as character direction — "90s anime noir" names not just a look but a type of person. Write one dominant light source per shot. Establish the character's internal condition before the external threat; internal stakes and external threat should be separate.
- Build 15-second character studies as micro-escalation sequences — six beats, six reveals, one location. Use camera proximity as the escalation axis. Ground every VFX in physics (reflect in rain, create distortion waves) so the technology reads as real rather than as an artifact.
Browse the full Scenic sci-fi and cyberpunk video gallery or read how to write Seedance 2 prompts for the complete cinematic prompting guide.
