Fight choreography fails in AI video for the same reason minimalist prompts fail: the model defaults to a generic register. "Martial arts fight scene" produces a recognizable output — two figures exchanging punches in a neutral space, speed-ramp on impact, slow-motion finish — because that's what the training data's most common fight clips look like. What separates a compelling fight sequence from that default is environment specificity, camera grammar tied to a specific cinematic tradition, and structural design that gives the fight a beginning, a middle, and an end at the prompt level.
The five prompts below each solve this differently. The K-drama rooftop uses wet-surface physics to build the cinematographic light conditions before a single punch lands. The hair salon fight makes the environment a weapon — the architecture breaks, mirrors shatter, the space itself participates in the choreography. The Rio capoeira prompt uses drone altitude as a narrative variable, starting with fighters as tiny geological features and closing in as the exchange intensifies. The ancient village karate prompt wraps explosive action inside a three-beat calm → chaos → calm arc that converts a fight clip into a complete character statement. And the cramped 80s garage uses spatial constraint as the choreographic generator — the limited space forces improvised-object problem-solving that period cinema grammar then reads as authenticity.
1. The K-drama rooftop — wet surface as cinematic light system
See the full prompt on scenic.sh →
"A cinematic, high-energy fight sequence on a rainy overcast rooftop in a Korean city, inspired by intense K-drama action. A fierce female high school student single-handedly takes on a group of male students in a brutal, realistic martial arts brawl."
Why this works: At 552 likes, this prompt demonstrates that environment specification isn't mood description — it's a physics call. "Rainy overcast rooftop" is not atmosphere; it's a specific light-multiplication event. Wet rooftop surface + overcast gray sky = the rooftop becomes a second luminous layer, reflecting the sky upward and softening the fighters' undersides. "Moody gray lighting" alone gives Seedance permission to light however it interprets moody gray. "Wet rooftop reflections" plus overcast specifies the mechanism: roof-level specular reflection is the fill source, not a placed light.
The camera grammar is equally precise: "handheld tracking shots, low-angle action perspectives, tight close-ups of intense expressions." This is the K-drama fight grammar in sequence — establish the space and wet terrain (tracking), enter the body-level of the fight (low angle), commit to the face and consequence (close-up). Each move pushes the viewer closer to the physical and emotional stakes. "Ultra-realistic live-action style, 4K, film-quality visuals" anchors Seedance to the photorealistic register, preventing the default action-cinema stylization that adds speed lines and impact glow.
Takeaway: Specify rain as a light-multiplication mechanism (wet surface + sky color = second luminous ground layer) rather than mood — the physics produce the photographic result, and "moody" alone doesn't.
2. The hair salon fight — environment as structural choreography
See the full prompt on scenic.sh →
"A dynamic, cinematic action sequence in a modern upscale hair salon at night. She shatters large mirrors and glass panels in dramatic slow-motion explosions of shards, throws attackers into salon chairs and shelves, and creates chaos as hair products, equipment, and debris scatter."
Why this works: At 410 likes, this prompt solves the core problem of fight scene design: making the environment a participant rather than a backdrop. "Shatters large mirrors and glass panels," "throws attackers into salon chairs and shelves," "hair products, equipment, and debris scatter" — each of these is an environmental element being consumed by the fight. The salon doesn't surround the action; it becomes the action's material.
Mirror shattering is specifically effective: a hair salon is one of the only spaces with 360° reflective surface coverage, which means a fighter surrounded by mirrors appears everywhere at once during the wide shot. The slow-motion glass-shard sequence, standard in action cinema, carries an additional register inside a salon — beauty tools and reflective surfaces transforming into kinetic projectiles. Character contrast does the work before a punch lands: "white cropped blazer, white crop top, and white shorts" against "multiple aggressive male attackers dressed in black tactical gear" — the fight is readable as color geometry from the establishing shot.
The camera follows the destruction: "dynamic handheld shots, pans, and dramatic angles" move with whatever the fight is breaking. The sequence ends on a character beat — she "calmly looks directly forward, runs a hand through her wavy blonde hair, and gazes at her reflection in a shattered mirror" — which uses the salon's central object (the mirror, now broken) as the closing image.
Takeaway: Build environmental destruction into the fight sequence as scene structure, not special effect — name the specific elements that will break and assign them to choreography beats; the environment becomes the fight's visual vocabulary.
3. The Rio capoeira drone — altitude as narrative variable
See the full prompt on scenic.sh →
"One continuous drone shot of 2 men fighting Capoeira on a mountain top platform overlooking the Rio Favelas. The drone opens on a high aerial view with fighters tiny within the massive landscape. The drone begins tracing a wide oval path, then banks inward and begins spiraling down."
Why this works: At 22 likes, this prompt demonstrates that the camera's altitude change is the narrative arc. The shot begins with fighters "tiny within the massive landscape" — at extreme wide, two humans become geological features, small relative to the mountain, the cliff platform, and the favelas layered below. As the drone descends through its spiral path and tightening orbit, the scale relationship inverts: by the exchange sequence, the fighters fill the frame and the city is peripheral.
Capoeira is specifically drone-legible: the style is built around circular movement, high leg arcs, and inversions — all of which read more clearly from above than from eye level. The "circular concrete fighting platform perched on the edge of a cliff" gives the drone a compositional anchor (the platform rim as a visual frame within the frame) that makes the orbiting motion coherent. As the drone descends and tightens its orbit during the exchange, the closeness matches the choreography's intensity — wide when the fighters circle in ginga, close during the kick exchanges.
The shot resolves with the fighters resetting to ginga as the drone "pulls slightly wider to reconnect the men to the circular platform, the cliff edge, the favela below, and the surrounding mountains." The drone's movement traces the same arc as the fight's emotional arc: vast → intimate → vast. The city is the context the camera relationship controls, not background.
Takeaway: For aerial fight documentation, altitude change is the narrative tool — start wide enough that fighters read as small within the landscape, descend to intimate scale during the exchange, pull back for resolution; the drone's flight path maps directly to the fight's emotional arc.
4. The ancient village tea arc — three-beat narrative structure
See the full prompt on scenic.sh →
"A female lone martial artist in an ancient Chinese village, interrupted while drinking tea, fights 10–20 attackers. Tone: calm → explosive precision → slow-motion control → calm resolution. Final scene: silence returns. She calmly sits back at the wooden table, picks up her tea, and takes a composed sip."
Why this works: At 24 likes, this prompt builds narrative closure into the action sequence at the prompt level — most fight prompts stop at the climax and never land. The four-beat arc is explicit in the prompt's own tone description: "calm → explosive precision → slow-motion control → calm resolution." The fight is structurally bracketed by the same behavior: drinking tea. The attacker interrupts; the fight is the interruption; returning to tea is the statement.
The overhead spinning camera shot during the fight does something most fight sequences miss: "mid-fight overhead spinning camera shot shows her white gi flowing dynamically, enemies pushed outward in a radial shock pattern." From above, the fight becomes geometry — the circular pattern of bodies radiating from a single center. The white gi against the distributed group converts the action frame into a legible abstract composition. The slow-motion moment follows: "brief slow-motion moment during a mid-air clash, emphasizing balance and precision." Slow motion is earned at the geometric peak.
The resolution beat — "silence returns, she calmly sits back at the wooden table, picks up her tea" — is what converts a fight clip into a character statement. It communicates that the fight was below her engagement threshold: an interruption resolved and set aside. Mastery expressed through return to ordinary behavior.
Takeaway: Build a three-beat narrative arc (calm → violent disruption → calm return to the opening action) for short fight sequences — the closing beat is what makes the fight a statement about character rather than a display of action.
5. The 80s garage — spatial constraint as choreographic generator
See the full prompt on scenic.sh →
"A fast-paced kung fu fight in a cramped garage, 80s action cinema style, grounded and physical. One protagonist faces 3 attackers, using improvised objects (chair and ladder) with fluid movement."
Why this works: This prompt demonstrates that the environment's physical constraints generate the choreography rather than containing it. "Cramped garage" removes the option of wide spinning kicks and extended acrobatic sequences — there isn't room. "Using improvised objects (chair and ladder)" follows directly from that constraint: in a real cramped space, you work with what's there. The problem-solving that spatial limitation forces reads as physical authenticity.
"80s action cinema style, grounded and physical" activates a specific cinematographic vocabulary: practical lighting (fluorescent tubes, work lights), tight two-shots and close-ups rather than wide establishing shots, physical contact over wire work or CG enhancement. The camera grammar of the period was built to sell impact — cut on the hit, hold the reaction, don't linger on the recovery. "Grounded and physical" bans the two defaults that break period authenticity: wire-assisted height and digital-era speed ramping.
"One protagonist faces 3 attackers" sets the dramatic context before the fight begins. The 3-vs-1 ratio is the stake that the cramped space, the improvised objects, and the 80s camera grammar must then carry. Each choice reinforces the others: garage → no room to retreat → use the chair → 80s camera holds on the improvisation.
Takeaway: Define the environment's physical constraints first — "cramped garage" generates the problem the choreography must solve, and problem-solving forced by a real space reads as physical authenticity that no amount of technique vocabulary can replicate.
Martial arts fight scene prompt cheat sheet
What these five Seedance fight prompts share:
- Environment as structural participant, not backdrop — wet rooftop physics multiply the light; salon mirrors become weapons; the cliff platform gives the drone a compositional anchor; tea cups bookend the arc; garage walls force improvisation. In each prompt, the space participates in the choreography rather than surrounding it.
- Camera grammar tied to a cinematic tradition — K-drama handheld tracking → low angle → close-up; 80s action two-shot → cut on impact; drone descent arc matched to emotional intensity. Naming the cinematic tradition ("K-drama," "80s action cinema style") activates a complete camera vocabulary that would take many individual instructions to replicate.
- Narrative arc, not sequence — the village tea prompt makes this explicit, but every prompt here has a beginning, middle, and end built into the prompt itself. Fight sequences without arc are clip demonstrations; fight sequences with arc are character films.
- Constraint as choreographic generator — spatial limits (cramped garage), environmental physics (wet surface), platform geometry (circular concrete edge) all force the fight to solve specific physical problems. Constraint generates distinctiveness; open space generates generic action.
- Specific material over style adjectives — "wet rooftop reflections," "smashed mirrors and glass panels," "chair and ladder," "white gi," "circular concrete platform." Material specificity tells Seedance what to render; style adjectives tell it how to feel — material specification generates the style as a byproduct.
→ Browse the Sports & Athletic Performance gallery on scenic.sh for more physically grounded prompts
→ For choreography and dramatic scene techniques, see 5 Seedance Action Video Prompts
→ For the complete Seedance prompt technique guide: How to Write Seedance 2 Prompts
