Short-form anime and cinematic AI video — original series, contest work, and commercial-style productions. Every project comes with the prompts behind it.
Selected Work
A few favorites — contest winners and the pieces I'm proudest of.
Portfolio
Full body of work — original series, contest entries, and commercial-style productions.
See more of my work on social media.
Prompt Vault
The exact prompts behind each production — filter by style or tool, copy, and go.
Art Style — Character Sheet & Background (GPT Image 2)
Bold, flat, stylized illustration with hard-edged cel-shading — large flat color blocks with angular, solid-colored shadows. No soft gradients, no painterly texture, no photorealism. Sharp, clean silhouettes with minimal outlines and confident, graphic character proportions. Style reference: high-end flat character illustration with a vivid, saturated color palette. Masterpiece quality, 4K resolution.
Video Prompt — Clip 1 (15s) — Seedance 2.0 / 2.5
Create a 15-second animated urban parkour chase sequence. Use @image1 as the environment reference, @image2 as the female pursuer reference, and @image3 as the mysterious male runner reference.
Visual style: bold, flat, stylized illustration with hard-edged cel-shading, large flat color blocks, angular shadows, vivid saturated colors, and sharp, clean silhouettes. No gradients, no painterly textures, no photorealism. High-end animated film quality.
The entire animation is ONE CONTINUOUS SHOT — no cuts, no transitions, no teleporting camera. Smooth 120fps appearance, with action pacing roughly 1.5x faster than normal.
The city is alive with pedestrians, cafes, apartment balconies, fire escapes, utility cables, rooftop structures, water tanks, and connected rooftops. People react naturally as the chase passes through.
00–05 seconds: Camera starts low behind @image3 as he sprints through a narrow alley, vaulting over crates, pushing through a crowded passage, and sliding beneath a partially closed metal shutter. The camera tracks aggressively just a few feet behind him, whipping around corners and obstacles. As he bursts out of the alley and reaches an outdoor staircase, the camera swings around to reveal @image2 emerging from the same route moments later, chasing him at full speed.
05–10 seconds: The chase climbs upward through fire escapes and maintenance stairways. @image3 reaches each platform first, clearing railings and obstacles in one fluid motion. @image2 follows the exact path, gaining ground with every movement. The camera becomes increasingly dynamic — rapid side tracking, aggressive low-angle movement, smooth orbiting around stair structures, quick elevation changes, and fast directional shifts, all connected in one seamless shot. The distance between them noticeably shrinks.
10–13 seconds: The chase reaches the connected rooftops. @image3 launches across narrow rooftop gaps, racing past water tanks, billboards, and ventilation units, with @image2 close behind. The camera performs a fast orbit around rooftop obstacles while tracking both runners. For the first time, @image2 comes within arm's reach of @image3.
13–15 seconds: @image3 makes one final desperate rooftop jump. The camera follows through the jump and performs a fast rotational movement during landing. As he recovers, @image2 lands directly behind him, nearly grabbing the back of his jacket. The camera swings to the front while moving backward, revealing his brief moment of panic as he realizes she is almost catching him. The chase continues at full speed toward the next rooftop as the video ends.
Camera style: Aggressive cinematic chase camera — rapid tracking, dynamic orbiting, whip-like directional changes, low-angle movement, side tracking, dramatic elevation shifts, and fast rotations, while remaining spatially readable and consistent throughout.
Audio: High-energy parkour chase music with driving percussion and rhythmic electronic elements, kept lower than native audio. Footsteps, breathing, clothing movement, rooftop impacts, metal staircases, crowd reactions, and city ambience remain clearly audible and dominant.
Video Prompt — Clip 2 (15s) — Seedance 2.0 / 2.5
Create a 15-second continuation of the previous rooftop chase. Use @image1 as the exact starting frame reference. Use @image2 as the environment reference, @image3 as the female pursuer reference, and @image4 as the male runner reference.
Visual style: bold, flat, stylized illustration with hard-edged cel-shading, large flat color blocks, angular shadows, vivid saturated colors, and sharp, clean silhouettes. No gradients, no painterly textures, no photorealism. High-end animated film quality.
The entire animation is ONE CONTINUOUS SHOT — no cuts, no transitions, no teleporting camera. Smooth 120fps appearance, with action pacing roughly 1.5x faster than normal.
The entire sequence takes place across connected rooftops filled with water tanks, ventilation units, maintenance walkways, utility cables, billboards, rooftop structures, and open skyline views.
00–03 seconds: Start exactly from @image1. The woman is almost grabbing the man's jacket. The man suddenly twists his body and slips free at the last second, and the woman stumbles forward slightly from the missed grab. The camera stays extremely close to both characters, capturing the near-capture moment. The man immediately accelerates into a sprint while the woman quickly regains balance and continues chasing.
03–07 seconds: The chase continues across connected rooftops. The man vaults over ventilation units, cuts sharply around rooftop structures, and uses efficient parkour movements to gain speed. The woman follows aggressively, taking the same route only moments later. The camera becomes highly dynamic — rapid side tracking, aggressive low-angle movement, fast forward pushes, smooth orbiting around rooftop obstacles, quick directional changes, and dramatic perspective shifts, while maintaining spatial clarity.
07–10 seconds: The man begins creating real distance for the first time. He clears a larger rooftop gap and lands smoothly, while the woman follows successfully but loses valuable momentum during her landing. The camera performs a fast sweeping movement revealing the growing gap between them. The woman visibly pushes harder, but the man continues pulling away.
10–13 seconds: The rooftops become more open as the man races toward the far end of the rooftop district, and the distance increases rapidly. The camera alternates between tracking the man ahead and revealing the woman further behind in the same continuous movement. For the first time, it becomes clear she may not catch him.
13–15 seconds: The man disappears beyond the final rooftop structures and out of view. The woman slows to a stop near the edge of a rooftop as the camera smoothly swings around from behind her to the front. She bends forward slightly, breathing heavily, her expression showing pure frustration, disappointment, and exhaustion. She looks toward the direction where the man escaped, clearly annoyed that he got away, as the city skyline fills the background and wind moves her hair and clothing. The camera slowly pushes closer to her annoyed expression as the sequence ends.
Camera style: Aggressive cinematic chase camera — rapid tracking, dynamic orbiting, whip-like directional changes, low-angle movement, side tracking, dramatic elevation shifts, and fast rotations, while remaining spatially readable and consistent throughout.
Audio: High-energy parkour chase music with driving percussion and rhythmic electronic elements, kept lower than native audio. Footsteps, breathing, clothing movement, rooftop impacts, wind, distant city ambience, and environmental sounds remain clearly audible and dominant.
Masterpiece quality, expressive animation, seamless one-take cinematography, dynamic storytelling, rooftop parkour pursuit, 16:9.
Workflow
Open ChatGPT — or any platform running the GPT Image 2 model — and generate two character sheets, one for the pursuer and one for the runner being chased, along with a separate background image. Apply the art style prompt above to all three. Using GPT Image 2 specifically matters here, since other models will produce noticeably different results.
Open any AI video generation platform that supports Seedance and paste the Clip 1 prompt above. Replace @image1 with your chosen background, @image2 with the pursuer, and @image3 with the runner being chased.
Once Clip 1 finishes rendering, take a screenshot of its final frame. Paste the Clip 2 prompt above and replace @image1 with that screenshot, @image2 with the background, @image3 with the pursuer, and @image4 with the runner being chased.
Note: if you're using Seedance 2.5, which can already generate a full 30-second clip in one pass, feel free to adapt this prompt into a single 30-second version. It's split into two 15-second prompts here intentionally, to show how to stitch separate scenes into one continuous-feeling sequence — a technique worth knowing any time you're building a longer video.
Art Style — Character Sheet (Midjourney v8.1)
Semi-flat editorial character illustration, chunky painterly brushwork, angular geometric rendering, faceted color blocking, clean shape-based painting, simplified anatomy with slightly exaggerated proportions, oversized contemporary clothing, minimal facial detail, expressive silhouette, soft ambient lighting, muted neutral color palette with carefully placed saturated accents, crisp value separation, graphic composition, modern fashion concept art, premium stylized character design, elegant painterly finish, subtle visible brush strokes, clean background, highly readable shapes, dynamic but uncluttered presentation, premium illustration quality.
Video Prompt — 15s (Seedance 2.0 / 2.5)
Create a 15-second continuous cinematic parkour sequence using @image1 as the only character.
Art style: semi-flat editorial illustration, chunky painterly brushwork, faceted geometric color blocks, clean shape-based painting, simplified anatomy, slightly exaggerated proportions, oversized clothing, minimal facial detail, muted neutral palette with subtle warm accents, graphic silhouette, premium stylized concept art. 24fps, smooth animation.
Parkour pacing feels 5× faster than normal. One continuous shot — no cuts, no teleportation, no background music.
Environment: a lively traditional Asian market during the late afternoon, with a realistic, believable layout. The market consists of narrow pedestrian alleys, fruit stalls, vegetable stalls, small restaurants, wooden awnings, balconies, drainage pipes, maintenance ladders, tiled rooftops, utility poles, hanging fabrics, wooden beams, railings, window ledges, and connected market buildings. Everything feels authentic and naturally busy, and every object belongs logically in the environment — no random props, no unnecessary decorations, no floating objects, no oversized signs, no abandoned vehicles, no cars or trucks inside narrow pedestrian alleys, and no AI-generated visual clutter. The market should feel clean, readable, and naturally lived in.
00–03 seconds: The camera begins low behind the character, as if another runner is chasing him through the market. He sprints between pedestrians, speed-vaults over an empty wooden handcart parked beside the walkway, runs briefly along a low brick wall, grabs a wooden support beam, and pulls himself onto the structural frame above a market awning. Without stopping, he continues running above the stalls. The camera aggressively chases him, weaving naturally through hanging fabrics, wooden posts, and market structures with realistic handheld movement.
03–07 seconds: Still one continuous shot, the character runs only along reinforced wooden beams and structural edges above the awnings. He performs a kong vault over empty transport crates stacked beside a building wall, climbs a fixed drainage pipe, and transitions smoothly onto a low tiled rooftop. He runs across rooftops, ducks beneath hanging electrical cables, performs precise foot placement across narrow rooftop gaps, and drops onto another reinforced awning support using a controlled parkour landing. He never steps on merchandise, and never touches fruit, vegetables, or market goods. The camera naturally alternates between rear-quarter tracking, side tracking, and brief frontal tracking while passing behind poles, signs, and hanging banners.
07–12 seconds: The route becomes more vertical. The character grabs a balcony railing, swings onto a maintenance platform, runs up a fire escape, vaults over rooftop walls, slides beneath hanging laundry, and uses another wall run to reach a higher rooftop. He drops briefly onto the roof of an empty delivery cart positioned in a service lane before pushing onto another rooftop. Every movement feels realistic for an experienced traceur — no superhuman jumps, no impossible distances.
12–15 seconds: The rooftops gradually descend toward his favorite fruit shop. The character performs one final controlled rooftop drop, executes a smooth parkour roll, and walks into the fruit shop naturally. He stops, picks up one fresh red apple, takes one bite, and looks satisfied. The fruit seller smiles naturally. End.
Camera: the camera behaves like a professional handheld chase operator rather than a video-game camera, constantly searching for stronger compositions while maintaining one uninterrupted shot. It alternates naturally between rear-quarter tracking, side tracking, brief frontal tracking, low-angle pursuit, short crane rises, foreground occlusion, fast whip pans during direction changes, small framing imperfections, slight operator delays, and natural handheld inertia. The camera never stays directly behind the character for too long, avoids gameplay-style follow camera and unnecessary orbiting, and always preserves clear spatial continuity.
Character consistency: only one @image1 appears — no clones, no duplicates, no transformations.
Parkour rules: the character never steps on merchandise, never runs across fruit, never steps on vegetables, and never damages market stalls. Every parkour movement uses only architectural structures, including walls, railings, drainage pipes, balconies, stairs, wooden beams, rooftops, awning support frames, window ledges, maintenance platforms, and support columns. Every obstacle must feel naturally built into the market rather than randomly placed.
Audio: native environmental audio only — fast footsteps, wood impacts, vault impacts, fabric movement, crowd reactions, shop ambience, bicycle bells, distant conversations, wind, and an apple bite at the ending.
Avoid: stepping on fruit, stepping on vegetables, breaking stalls, cars inside narrow alleys, random props, floating objects, visual clutter, illogical obstacle placement, video-game style movement, superhuman jumps, and unnecessary cinematic effects.
Workflow
Create a character sheet using Midjourney v8.1 with the art style prompt above.
Copy the video prompt above and replace @image1 with your character. Use Seedance 2.0 or Seedance 2.5.
Val Gregory
Independent AI content creator based in Indonesia, working across short-form anime, cinematic video, commercial ads, and more.
Experienced in producing content for commercials, brands, and projects spanning any concept or art style. Has entered more than 40 contests, with 21 wins to date.
If you're a brand, project, or individual interested in collaborating — or would like a custom video made — feel free to get in touch.
CPP
OpenArtImagineArtOii OiiFlova AIDreamina AIPixVerseKling AI