Model Strengths — Quick Reference
Google Veo 3.1
Best cinematic quality. Broadcast-ready color science. Exceptional wide shots, lighting, weather, fog, rain. 24fps film look. Our workhorse for environmental beauty shots.
Kling 3.0
Best motion quality + multi-shot narratives. Start/end frame control. Character consistency across scenes. Physics-aware. Up to 15sec native. Our go-to for action sequences.
Sora 2
Best physics accuracy + temporal consistency. Objects behave realistically. Strong depth/parallax for drone shots. Camera language control (lens, motion). Hero close-ups.
Seedance 2.0
Best multimodal control — accepts images, video, audio as reference inputs. Character consistency across multiple scenes. Ideal for style-locked sequences.
Higgsfield Soul
Best photorealistic still images. Use for key frame generation and storyboard images BEFORE animating with video models. Locks composition and tone.
Cinema Studio
Start + end frame control. Multi-shot narrative logic. Timeline editing. Use for assembling sequences and ensuring shot-to-shot continuity. The editing backbone.
Flux Kontext
Image editing and style transfer. Use for refining key frames — adding HUD overlays, adjusting lighting, fine-tuning details before animation.
Nano Banana Pro
Gemini 3.0 backbone. Native 4K, reasoning-guided generation. Flawless text rendering (ideal for HUD graphics). Character consistency across images. Multi-image fusion (up to 8 refs). ~10sec generation. Our top pick for key frames with text overlays.
Kling O1
World's first unified video+edit model. Element Library locks character identity across shots. Semantic editing via natural language (remove objects, swap backgrounds). Motion transfer from reference videos. Start/end frame transitions. 3–10 sec clips. Our VFX and continuity editor.
Wan 2.5
Native audio generation — dialogue, ambient sound, music synced to visuals in one pass. 1080p HD at 24fps, up to 10 sec. Video-to-video with lip-sync. Camera control presets (dolly, FPV, bullet time). Our audio-visual integration engine.
Seedream 4.0 / 4.5
ByteDance's ultra-realistic image model. Fashion-grade cinematic lighting, soft textures, photo-studio composition. Multi-image editing with identity preservation. 4K output. Strong prompt adherence. Our editorial realism engine for stills.
MiniMax Hailuo 02
#2 globally on Artificial Analysis. Native 1080p. Extreme physics mastery — stunts, acrobatics, kinetic camera work. NCR architecture for 2.5x efficiency. Best for high-motion, high-energy action shots with believable momentum.
Banana Placement / Product Placement
Precision object placement in generated scenes. Draw-to-video with visual storyboards instead of prompts. Works with Veo 3 and Hailuo 02. Use for placing props, logos, or objects into specific frame positions.
Grok Image / GPT Image
Backup image generators for concept exploration. Quick idea testing. Use if Soul / Nano Banana Pro / Seedream don't nail a specific look.
Recommended Production Chain Per Shot
1
Generate the key frame as a still image
Nano Banana Pro / Soul / Seedream 4.5 / GPT Image
2
Refine composition, add HUD/overlay details
Nano Banana Pro (text) / Flux Kontext / Seedream
3
Animate the image into video with the best-fit model
Veo / Kling / Sora / Seedance / Hailuo 02 / Wan 2.5
4
Edit, refine, VFX, and ensure continuity across shots
Kling O1 / Cinema Studio / Banana Placement
5
Final edit, sound design, color grade, watermark
External Editor + Higgsfield
CAR IDENTITY SPEC
Matte black, low aggressive prototype GT, 2035-era. Carbon fiber body panels and accents. No brand logos, no sponsor markings (instant DQ). Car number "12" in small white on the door. Single cyan-lit AEGIS module panel visible on dashboard interior. No external LED decorations (rivals have them, Kira doesn't). When AEGIS is killed in Act 4, all cyan lights on the car go dark. Headlights are sharp, white, functional.
Vehicle Reference Prompts — Nano Banana Pro / Seedream 4.5
Generate 8-10 images. Feed the best into Kling O1 Element Library or use as reference images for Veo/Sora/Kling image-to-video. Keep the same session open for consistency.
V1 — Three-Quarter Front (Hero Shot)
Cinematic three-quarter front view of a matte black prototype racing car parked on wet asphalt at night. Low, aggressive, 2035-era GT shape with sculpted carbon fiber body panels. The car sits low to the ground with wide haunches and minimal ground clearance. No brand logos, no sponsor decals, no text anywhere on the exterior. Small white number "12" on the driver's door. No LED strips or decorative lighting on the exterior body. The only light comes from sharp white headlights reflecting off wet ground. A faint cyan glow is visible through the cockpit windshield from the dashboard AEGIS module. Dark forested background, rain falling. Photorealistic, 4K, cinematic, anamorphic, shallow depth of field. The car should look purposeful, stripped, analog compared to its environment.
V2 — Full Side Profile
Cinematic full side profile of a matte black prototype racing car on a wet race track at night. Low, aggressive silhouette, 2035-era GT proportions, long hood, short rear overhang, sculpted carbon fiber panels. No logos, no sponsor markings, no decorative LEDs. Small white "12" on the door. Headlights off. The car is a dark silhouette against faint ambient light reflecting off wet asphalt. Rain falling. The car's shape should read clearly as a purpose-built racing machine, minimal and stripped of anything cosmetic. Photorealistic, 4K, cinematic, high contrast, anamorphic.
V3 — Rear Three-Quarter (Drift Angle)
Cinematic rear three-quarter view of a matte black prototype racing car on wet asphalt at night. The rear end is wide, muscular, with a subtle integrated rear wing and quad exhaust tips. Carbon fiber diffuser visible underneath. Matte black finish catches faint reflections from rain on the ground. Red tail lights glowing. No logos, no brand markings, no LED decorations. Small white "1" visible on the rear fender. Tire spray mist hanging in the air behind the car. Dark forest background. This is the angle the audience will see during the The Fold drift. Photorealistic, 4K, cinematic, anamorphic, dramatic lighting.
V4 — Cockpit Interior (AEGIS Active)
Cinematic interior view of a 2035-era prototype racing car cockpit. The camera is positioned from the passenger side looking toward the driver's seat. Carbon fiber surfaces everywhere, stripped and functional. A flat-bottom carbon fiber steering wheel with minimal buttons. A small digital display cluster showing faint cyan data readouts. The AEGIS module is a rectangular panel on the center dashboard emitting a soft cyan glow. A physical toggle switch with a red metal safety guard is mounted on the center console, labeled "MANUAL" in small engraved text. The cockpit is tight, purposeful, no comfort features. Faint rain visible through the windshield. Lighting: cyan glow from AEGIS module and dashboard instruments, warm amber from a secondary display. Photorealistic, 4K, shallow depth of field, cinematic.
V5 — Cockpit Interior (AEGIS Dead — Act 4)
Cinematic interior view of the same 2035-era prototype racing car cockpit, but now every screen is BLACK. Every digital display is dead. The AEGIS module panel on the dashboard is completely dark, no cyan glow. The only light source is faint moonlight and distant ambient light coming through the windshield, barely illuminating the carbon fiber surfaces. The physical MANUAL toggle switch is now flipped to the ON position, its red safety guard open. Rain streams down the windshield. The cockpit feels raw, analog, unplugged. This is what it looks like when you kill every electronic system in a modern race car and drive blind. Photorealistic, 4K, near-total darkness, cinematic, anamorphic.
V6 — On Track, Headlights in Rain (Act 4 Mood)
Cinematic wide shot of a matte black prototype racing car driving alone on a rain-soaked race track at night. No digital displays visible on the car, no cyan glow, no LED decorations. The only light sources are the car's sharp white headlights cutting two beams through heavy rain. The wet track surface reflects the headlight beams. Dark forest lines both sides of the narrow track. The car is small in the frame, surrounded by darkness and weather. No other cars visible. Complete isolation. This is a machine and a human, nothing else. Photorealistic, 4K, ultra-wide 2.39:1 aspect ratio, cinematic, film noir atmosphere, anamorphic.
After generating car references: Upload the best 6-8 into Kling O1 Element Library. This gives Kling "memory" of the car across all subsequent video generations, the same way Soul ID locks Kira's face. For Veo and Sora (which don't have an Element Library), attach the best car reference image as an image-to-video input alongside your scene prompt. For Seedance, use multi-reference mode with both a Kira ref AND a car ref.
Each shot card below now includes a VIDEO PROMPT section. These are structured for Higgsfield's video models with these components:
Veo 3.1: Describe the scene cinematically. Include camera movement, lighting, atmosphere, and mood. Attach a key frame image as input. Veo reads the image and extends it into motion. Keep prompts descriptive and atmospheric.
Sora 2: Similar to Veo but emphasize physics behavior (acceleration, weight transfer, water dynamics). Sora responds well to lens language ("anamorphic," "shallow depth of field," "35mm"). Attach key frame as start image.
Kling 3.0: Use start/end frame control. Generate or select two key frames (beginning and end of the shot) and let Kling interpolate the motion between them. Add a motion prompt describing what happens. Kling excels when you define the transformation.
Wan 2.5: Same as Veo/Sora but add audio direction in the prompt. Wan generates sound natively. Describe what you HEAR as well as what you see.
Hailuo 02: Excels at physics-heavy motion. Be specific about forces, momentum, and kinetic energy in the prompt. Attach key frame as input.
REFERENCE IMAGES: For every shot, attach: (1) Kira Soul ID reference, (2) Car reference from Element Library or as attached image, (3) The key frame still you generated for that shot. The more visual references, the more consistent the output.
Shot 01
Kira's Eyes — The Last Driver
Ironwood Raceway Front Straight — Grid Position
Hero Shot
Extreme close-up of Kira's eyes behind her visor. Grid lights reflected in the glass — cyan and red neon. Rain droplets on the helmet surface. Her breathing fogs the visor slightly. Slow, deliberate. This is the face of the film.
ECU
Dolly In (slow)
Shallow DOF
3 sec
Generate the hero portrait still. Soul excels at photorealistic faces with precise lighting. Lock the visor reflections, skin detail, and helmet texture here.
Native 4K with reasoning-guided rendering. Can generate the hero portrait AND render visor HUD text ("MANUAL MODE") in a single pass — no compositing needed. Character consistency across re-generations. Test both Soul and NBP, use whichever locks Kira's face best.
Sora's tight crop and close-up strength is perfect for faces. It handles subtle micro-movements (breathing, eye shifts, visor fog) with the best temporal consistency. Slow dolly in.
Higgsfield Prompt
Cinematic extreme close-up of a woman race car driver's eyes behind a racing helmet visor, neon blue and red grid lights reflected in the visor glass, rain droplets on helmet, dramatic shallow depth of field, anamorphic lens flare, dark background, futuristic 2035 motorsport, her breathing creates slight fog on the visor, the camera slowly dollies in, photorealistic, tense atmosphere
Video Generation Prompt — Sora 2
Extreme close-up of a woman's eyes behind a dark racing helmet visor. Very slow dolly in. She blinks once, slowly. A faint breath fogs the inside of the visor then clears. Rain droplets slide down the exterior visor surface. Cyan and red neon lights reflected in the visor glass pulse gently. Movement is minimal, hypnotic, like a heartbeat. Anamorphic lens, shallow depth of field, photorealistic, cinematic, 24fps film grain.
ATTACH: Kira Soul ID ref + Prompt 05 key frame (eyes/visor still)
Shot 02
The Futuristic Grid at Ironwood Raceway
Ironwood Raceway Front Straight — 3,000ft — Wide Establishing
Hero Shot — World Build
Wide establishing shot: futuristic race cars on Ironwood's front straight at night. Holographic position displays hover above each car. LED track markers glow through mist. Virginia's rolling hills silhouetted against a dark sky. Wet asphalt reflecting everything. This shot sells the WORLD.
Wide
Crane Down
4 sec
Veo dominates wide shots with global illumination, weather effects, and environmental scale. It handles fog, rain, wet reflections, and depth of field across large scenes better than any other model. The 24fps film look gives it that cinematic gravity. This is Veo's sweet spot.
Higgsfield Prompt
Cinematic wide shot of a futuristic night race grid, sleek race cars with glowing LED accents lined up on a long straight, holographic position displays hovering above each car, wet asphalt reflecting cyan and red neon track lights, mist rolling across the track, forested rolling hills silhouetted against dark sky, crane shot moving slowly downward, atmospheric rain, volumetric lighting, 2035 motorsport, photorealistic
Video Generation Prompt — Veo 3.1
Wide establishing shot of a futuristic racing grid at night. Camera starts high (crane position) and slowly descends toward track level over 4 seconds. Multiple sleek race cars sit on a wet straight, holographic displays hovering above each. Mist rolls across the track surface. Rain falls steadily. Cyan and red neon track markers reflect in the wet asphalt. Forested hills silhouetted against a dark overcast sky in the background. The camera descent reveals the scale of the grid. Volumetric light through rain. Photorealistic, cinematic, 24fps, anamorphic, 2035 motorsport.
ATTACH: Car ref (V1 three-quarter front) + Prompt 13 key frame (grid environment still)
Shot 03
The MANUAL Toggle
Cockpit Interior
Key Narrative Moment
Insert shot: Kira's gloved finger flips a physical toggle switch labeled "MANUAL" in green. The surrounding HUD elements dim as AEGIS goes to advisory mode. A rebellious act that defines her character in one gesture.
Insert / Macro
Static
2 sec
Kling's physics-aware motion handles mechanical interactions (finger on switch, toggle flipping) with the most natural movement. Start/end frame control lets you define the "before" (switch off) and "after" (switch on, HUD dims). Precise control over a small, detail-rich frame.
Higgsfield Prompt
Cinematic macro close-up of a gloved hand flipping a physical toggle switch on a futuristic race car steering wheel, the switch is labeled MANUAL in green text, surrounding holographic HUD elements dim and fade as the switch engages, dark cockpit interior with subtle blue ambient light, carbon fiber texture visible, photorealistic detail, shallow depth of field, 2035 motorsport technology
Video Generation Prompt — Kling 3.0 (Start/End Frame)
START FRAME: Macro close-up of a toggle switch labeled "MANUAL" in the OFF position. Red safety guard is closed over the switch. A gloved hand rests near it. Cyan HUD glow illuminates the carbon fiber console. END FRAME: The safety guard is flipped up, the toggle is thrown to ON. The cyan HUD glow has dimmed to near-darkness. The gloved hand is pulling away from the switch. MOTION: The hand lifts the red guard with a deliberate mechanical motion, then flips the toggle with a decisive click. The surrounding cyan light fades as the switch engages. 2 seconds. Photorealistic, macro detail, shallow depth of field.
ATTACH: Prompt 10 key frame (switch detail still) + Car cockpit ref (V4)
Shot 04
Lights Out — Launch
Ironwood Raceway Front Straight — Race Start
Key Action
The grid lights go out. Explosive launch. Cars rocket forward with tire smoke and spray. The front straight of Ironwood Raceway stretches ahead into darkness. Voiceover: "Tonight, Kira Vasquez races for more than a championship."
Low Angle Wide
Static → Rapid Pan
3 sec
Sora's physics engine handles the launch dynamics — tire smoke, water spray, acceleration forces — with the most convincing real-world behavior. Objects under force behave naturally. The sudden shift from static to violent motion is Sora's strength.
Higgsfield Prompt
Cinematic low-angle wide shot of futuristic race cars launching from a grid at night, tire smoke and water spray erupting, LED track lights blur as cars accelerate violently forward, the camera shakes slightly from the engine vibrations, dramatic lighting from overhead gantry, wet asphalt, photorealistic physics, 2035 motorsport, explosive energy
Video Generation Prompt — Sora 2
Low-angle wide shot at track level. Grid lights above go out in sequence: red, red, red, red, lights out. Instant explosion of motion: multiple race cars launch forward simultaneously. Tire smoke and water spray erupt behind each car. The camera shakes subtly from engine vibration and the blast of displaced air. LED track lights blur into streaks as the cars rocket away from camera. A matte black car (number 12, no decorative LEDs) pulls ahead of the pack. Wet asphalt, photorealistic physics, violent acceleration, 3 seconds, 24fps, cinematic.
ATTACH: Car ref (V1 front three-quarter) + Grid key frame from Prompt 13
Shot 05
The Spine — Full Flight
Ironwood Raceway Turns 7–9 — The Spine — 130+ mph
Hero Shot — Speed Showcase
FPV drone tracking shot following Kira through the iconic uphill esses. The car launches over crests, sparks fly from the undertray. LED track markers blur into light streaks. Other cars' AEGIS systems glow blue as they follow conservative lines — Kira cuts inside them all. Pure mastery.
FPV Drone
Tracking
8 sec
Veo handles the wide environment — night track, trees, elevation changes, weather — with unmatched cinematic lighting. The rolling hills and forest backdrop are Veo's specialty.
If you need this as a multi-shot sequence (approach → crest → landing), Kling's scene-based generation can handle the continuity. Test both and compare. Kling's motion smoothness may edge out Veo for pure speed feel.
Hailuo 02's extreme physics mastery makes it a dark horse here — it handles kinetic camera work, momentum, and high-energy action with believable weight. Native 1080p. Worth testing against Veo for the speed feel.
Higgsfield Prompt
Cinematic FPV drone tracking shot following a sleek futuristic race car through high-speed uphill S-curves at night, the car launches slightly airborne over a crest, orange sparks shower from the undertray, LED track markers blur into light streaks, rain beginning to fall, other cars visible behind with blue glowing AI displays, dark forested hillside background, extreme motion blur, photorealistic, intense speed, night racing, 2035
Video Generation Prompt — Veo 3.1 | ALT: Hailuo 02
FPV drone tracking shot, camera flying low behind and slightly above a matte black race car as it attacks a series of uphill S-curves at night. The car transitions left-right-left through three curves in rapid succession, climbing elevation with each turn. Over the crest, the car goes slightly airborne and orange sparks shower from the undertray on landing. LED track markers blur into streaks of cyan light. Rain intensifies. Other cars with blue-glowing AI displays are visible falling behind. Dark forested hillside on both sides. Extreme motion blur on background, the car stays sharp. 8 seconds, continuous shot, photorealistic, 24fps, anamorphic, pure speed.
ATTACH: Car ref (V3 rear three-quarter, the angle the drone will see) + Soul ID for any Kira cockpit cutaways
Shot 06
Overtake Montage
Ironwood Raceway Various — Speed Sequence
Supporting
Quick-cut montage: multiple angles of Kira overtaking AI-assisted cars. Cockpit POV, trackside, onboard rear camera. Each cut is 1-2 seconds. Sparks, rain spray, tire squeal implied. She's making it look effortless.
Multiple Angles
Quick Cut
6 sec total
Seedance's multimodal reference system is ideal for a montage — you can feed it reference images/video of the car from Shot 05 and get consistent style across multiple angle variations. Character + vehicle consistency across scenes is Seedance's defining advantage.
Video Generation Prompt — Seedance 2.0 (Multi-Reference)
Generate 3 separate 2-second clips for quick-cut montage:
CLIP A — Cockpit POV: Looking through windshield, a rival car with blue glowing AI displays is directly ahead. The matte black car pulls alongside and passes it. Rain on windshield. Cyan HUD data visible at edges of frame.
CLIP B — Trackside camera: The matte black car streaks past camera at high speed, followed closely by two rivals with blue LED accents. Water spray rooster-tails behind all three cars. Motion blur, night, wet track.
CLIP C — Onboard rear camera: Looking backward from the matte black car, showing rivals falling behind. Their blue AI displays shrink into the distance. Rain and tire spray fill the frame. Each clip photorealistic, 24fps, cinematic.
ATTACH: Car ref (V1, V2, V3 for multi-angle) + Kira Soul ID + any successful Shot 05 output as style reference
Shot 07
Ghostline — "Correction Applied"
Ironwood Raceway Turn 10 — Ghostline — Blind Crest
Hero Shot — The Turn
Cockpit POV: approaching Ghostline's blind crest. The steering wheel subtly moves on its own — just a degree. Kira's hands tighten. HUD flickers: "CORRECTION APPLIED." Her eyes shift from focus to suspicion. Everything changes here.
POV Cockpit
Static → Subtle Shake
4 sec
This shot requires precise mechanical motion (steering wheel moving autonomously) + subtle facial emotion shift. Kling handles both: physics-aware object motion and expressive character detail. Start/end frame control lets you define the exact "before/after" of the steering pull.
Use Flux to add/refine the "CORRECTION APPLIED" HUD text overlay on the key frame before animating. Clean typography on screen elements.
NBP can generate the cockpit key frame with "CORRECTION APPLIED" text rendered directly in-image — flawless text rendering means no compositing step. Test NBP vs Flux overlay approach.
Higgsfield Prompt
Cinematic cockpit POV of a futuristic race car approaching a blind hill crest at night, the steering wheel subtly rotates on its own, driver's gloved hands tighten on the wheel, a holographic HUD display flickers with a warning message in cyan text, the driver's eyes behind the visor shift from focused to alarmed, rain on the windscreen, dark track ahead with LED markers, tense atmosphere, photorealistic, 2035
Video Generation Prompt — Kling 3.0 (Start/End Frame)
START FRAME: Cockpit POV, looking through rain-streaked windshield at a dark uphill track approaching a crest. Gloved hands rest naturally on the steering wheel. Cyan HUD data visible at the edges, calm and nominal. END FRAME: Same POV but the steering wheel has rotated 3-5 degrees to the right ON ITS OWN. The driver's hands have tightened visibly, knuckles pressed against the gloves. The HUD now shows "CORRECTION APPLIED" in amber text. The driver's eyes (visible in reflection or via mirror) have shifted from relaxed to alert. MOTION: The wheel moves autonomously, slowly but unmistakably. The hand grip tightens in reaction. The HUD text flickers into view. Subtle camera shake as the car crests the hill. 4 seconds. Tense, deliberate, the first sign something is wrong.
ATTACH: Car cockpit ref (V4) + Kira Soul ID + Prompt 09 cockpit key frame
Shot 08
The Gate — Near-Spin
Ironwood Raceway Turn 11 — The Gate Corner
Key Action
Wide tracking shot: AEGIS brakes 50 meters too early into The Gate. Kira overrides but the car rotates — she catches it with opposite lock. Sparks shower off the curbing. The 4,000ft back straight opens up ahead as she gathers it.
Wide Tracking
Dolly
5 sec
The near-spin requires physics accuracy — weight transfer, tire slip, opposite lock correction. Sora's world-modeling engine handles rotational physics and momentum better than any other model. The car needs to behave like a real car losing and regaining grip.
Hailuo 02 is the only model capable of "extreme physics" like gymnastics-level motion. A car spin-and-catch is exactly its wheelhouse — momentum, weight shifts, tire slip. Test against Sora 2 and use whichever nails the slide.
Higgsfield Prompt
Cinematic wide tracking shot of a futuristic race car sliding sideways through a tight corner at night, the rear end breaks loose with tire smoke, the driver corrects with opposite steering lock, orange sparks fly from the car hitting the curb, a long dark straight opens ahead, wet asphalt, LED track markers, dramatic lighting, photorealistic physics, intense action, 2035 night racing
Video Generation Prompt — Sora 2 | ALT: Hailuo 02
Wide tracking shot, camera dollying alongside a matte black race car as it enters a fast corner at night. Midway through the corner the rear end snaps loose, the car rotates 30-40 degrees sideways. Tire smoke erupts from the rear wheels. The front wheels turn sharply into opposite lock as the driver catches the slide. Orange sparks shower from the left side as the car's undertray scrapes the inside curbing. The car gathers, straightens, and accelerates hard onto a long dark straight. Water spray hangs in the air behind it. The whole sequence happens fast, violent, then suddenly controlled. 5 seconds, photorealistic physics, weight transfer must look real, 24fps, cinematic, anamorphic.
ATTACH: Car ref (V2 side profile + V3 rear three-quarter) for the slide angle
Shot 09
Back Straight — Throttle Cut
Ironwood Raceway Back Straight — 4,000ft
Supporting
Cockpit shot: Kira accelerates down the back straight. At 190mph, AEGIS cuts the throttle without warning. The engine note drops. She finds the manual fuel bypass. Engine screams back to life. Quick, visceral, mechanical.
Cockpit Interior
Static → Handheld
4 sec
Interior cockpit with mechanical hand interactions (finding bypass switch). Kling's motion brush and start/end frame control can define the exact moment the throttle cuts and resumes. Scene-level control.
Video Generation Prompt — Kling 3.0 (Start/End) | ALT: Wan 2.5 (Audio)
START FRAME: Cockpit interior, Kira's hands on wheel, looking through windshield at a long dark straight stretching ahead. Speed blur through side windows. HUD shows nominal cyan data. Engine note is high and steady. END FRAME: Same cockpit but the digital dash has gone dark momentarily. Kira's right hand has left the wheel and is reaching toward a panel on the lower right console. MOTION: The car accelerates down the straight. Suddenly the engine note DROPS (throttle cut). Kira's body jerks forward slightly from deceleration. Her right hand leaves the wheel and searches the console. She finds a switch, flips it. The engine screams back to life. Her hand returns to the wheel. 4 seconds, fast-paced, visceral, mechanical. WAN 2.5 ALT: Include audio — high-pitched engine building RPM, sudden silence of throttle cut, cockpit wind noise, engine restarting with an aggressive bark.
ATTACH: Car cockpit ref (V4) + Kira Soul ID
Shot 10
HUD — "Relinquish Control"
Cockpit — Approaching The Plunge
Key Narrative
Cockpit close-up: the HUD turns red. AEGIS message fills the screen: "RELINQUISH CONTROL." The steering locks. Through the windscreen we see the wall rushing closer. This is the setup for the climax.
Cockpit Close-up
Static → Crash Zoom
3 sec
Generate the cockpit still with Soul. Use Flux Kontext to add the red "RELINQUISH CONTROL" HUD text overlay with precision. This shot is heavy on designed graphics.
NBP is the strongest choice here — it can render "RELINQUISH CONTROL" in red directly inside the cockpit image at 4K with perfect typography. Eliminates the Soul → Flux compositing step. Its reasoning engine understands spatial placement of text on screens.
Animate with start frame (normal cockpit) → end frame (red HUD, locked steering, wall approaching). Kling's frame control makes this transition precise and controlled.
Video Generation Prompt — Kling 3.0 (Start/End Frame)
START FRAME: Cockpit close-up, HUD displays showing amber warning data. Through the windshield, a downhill corner approach is visible. Hands on wheel, tension building. END FRAME: The entire HUD has turned aggressive RED. The text "RELINQUISH CONTROL" dominates the center display in large red letters. The steering wheel has locked, Kira's hands are fighting it. Through the windshield, a barrier is much closer. MOTION: The HUD color shifts rapidly from amber to red. Text elements multiply across the display. "RELINQUISH CONTROL" appears and pulses. The steering locks mid-turn. The camera begins a slow crash zoom toward the windshield and the approaching barrier. 3 seconds, escalating tension, the visual equivalent of a scream.
ATTACH: Car cockpit ref (V4) + NBP "RELINQUISH CONTROL" key frame
Shot 11
The Plunge — "No."
Ironwood Raceway Turn 14 — The Plunge — THE CLIMAX
★ THE HERO SHOT ★
The single most important shot in the film. Extreme close-up of Kira: calm, certain, defiant. She says "No." She reaches down and PULLS the breaker. Every screen in the cockpit goes BLACK. Complete darkness. The human takes over.
ECU Face
Crash Zoom → Static
4 sec
This is a PERFORMANCE shot. Sora's tight crop and temporal consistency make it the best for sustained close-up facial emotion. The transition from tension to calm resolve needs to read on her face. Sora handles this micro-expression range better than any model. Also: when the screens go black, Sora's lighting response will handle the sudden darkness shift naturally.
If you want her lips to actually say "No." with synced audio, Kling's lip-sync and dialogue engine is the industry leader. Test both: Sora for pure visual emotion, Kling for spoken word sync.
Wan 2.5 can generate this shot with native lip-synced dialogue + ambient audio in a single pass. Prompt the "No." as spoken dialogue + "mechanical THUNK of a breaker switch, then total silence." Could deliver the entire audio-visual moment without post-production.
Higgsfield Prompt
Cinematic extreme close-up of a woman race car driver's face behind her visor, illuminated by red warning lights from her HUD display, her expression shifts from tension to calm defiance, she speaks a single word, her hand reaches down out of frame, suddenly all light sources on her face go BLACK leaving only the faintest ambient glow, the transition from red-lit to total darkness is dramatic and sudden, photorealistic, anamorphic, intense emotional peak
Video Generation Prompt — Sora 2 | ALT: Kling 3.0 (Lip Sync) | ALT: Wan 2.5 (Full AV)
Extreme close-up of Kira's face behind her visor, bathed in harsh red HUD light. Her expression is tense, jaw clenched. Then something shifts: her jaw relaxes, her eyes steady, a calm certainty settles across her face. Her lips part and she says one word. Her right hand drops out of frame toward the center console. There is a physical mechanical THUNK sound. In ONE FRAME, every light source on her face dies. Red to absolute black. Instant. Not a fade, a severance. For 1-2 seconds, near-total darkness, only the faintest ambient glow on her cheekbones. Her breathing is the only sound. 4 seconds total. The most important 4 seconds in the film. Photorealistic, anamorphic, extreme facial detail. WAN 2.5 ALT: Generate with audio. Her voice says "No." clearly but quietly. Followed by the mechanical thunk of a breaker switch. Then absolute silence.
ATTACH: Kira Soul ID + Prompt 12 ("No." defiance still) + Prompt 08 (red helmet still for start frame lighting)
Shot 12
The Silent Drift — The Fold
Ironwood Raceway Turns 16–17 — The Fold — Into Front Straight
★ THE MONEY SHOT ★
The most beautiful frame in the film. Slow motion wide: Kira's car drifts through The Fold in total darkness — no HUD glow, no LED displays. Just the silhouette of the car, rain catching faint ambient light, tire smoke curling upward. She feeds onto the front straight. The only sound is rubber on wet asphalt.
Wide
Slow Motion
Static or Very Slow Dolly
5 sec
This shot is ALL atmosphere: darkness, rain, ambient light, tire smoke, silhouette. Veo's global illumination, weather simulation, and cinematic color science will make this frame a painting. The 24fps motion blur adds that dreamlike quality. No other model handles "dark beauty" like Veo. This is the shot you put on the poster.
Wan 2.5's native audio generation could produce this shot with built-in sound design — tires on wet asphalt, rain tapping, breathing — all synced to the visual. If the audio matches the mood, it saves an entire sound design pass. Prompt: "no music, only the sound of tires on wet asphalt and rain falling."
Higgsfield Prompt
Cinematic slow motion wide shot of a race car drifting through a dark downhill corner at night, no illuminated displays on the car, complete darkness except faint ambient moonlight catching rain drops and tire smoke rising from the rear wheels, the car is a dark silhouette against the faintest glow of a distant horizon, wet asphalt reflects nothing, ethereal and haunting atmosphere, extreme slow motion, photorealistic, film grain, the most beautiful and lonely frame in cinema, no artificial light sources
Video Generation Prompt — Veo 3.1 (THE MONEY SHOT) | ALT: Wan 2.5 (Audio)
Wide shot, slow motion. A matte black race car enters frame from the right, traveling fast. It turns into a tight downhill hairpin corner. The rear breaks loose into a controlled drift. The car slides through the apex at extreme angle, rear tires smoking, water spray arcing in a beautiful curtain behind it. There are NO digital lights on the car, no HUD glow, no LED displays. All electronic systems are dead. The only light sources are the car's white headlights cutting two beams through heavy rain, and faint moonlight filtering through clouds. The wet track surface catches the headlight beams. Dark Virginia forest lines both sides. Tire smoke curls upward, catching faint ambient light. The car gathers from the drift and feeds onto a straight, accelerating away from camera. 5 seconds, slow motion, photorealistic, 24fps, anamorphic, extreme atmosphere, film grain. This is liberation and danger. Silence and beauty. The most cinematic frame in the film. WAN 2.5 ALT: Audio only — tires hissing on wet asphalt, rain tapping gently, low unfiltered engine growl, faint breathing. No music. No electronic sounds. The quietest and most beautiful moment.
ATTACH: Car ref (V3 rear three-quarter + V6 on-track headlights) + Prompt 14 key frame (The Fold environment still)
Shot 13
Crossing the Line
Ironwood Raceway Front Straight — Finish
Key — Payoff
Wide shot: the car crosses the finish line on Ironwood's front straight. Minimal — no celebration, no crowd reaction. Just the car, the line, and the beginning of the end. Smash to black.
Wide
Static
3 sec
Match the visual style of Shot 02 (the grid). Same location, same wide framing, but now the grid is empty and the car is alone. Veo maintains environmental consistency. Book-ending the film with matching Veo shots gives it cohesion.
Video Generation Prompt — Veo 3.1
Wide static shot of a long, wet front straight at a race track at night. The same grid from Shot 02, but now empty, dark, abandoned. No other cars. No holographic displays. A single matte black race car enters from the far left of frame, traveling fast but not frantic. It crosses a simple white painted finish line on the track surface. A single timing light blinks green briefly as it passes. No fanfare, no celebration, no crowd. The car continues past the line and out of the right side of frame. The frame holds on the empty track for one beat. Smash to black. 3 seconds. The anticlimax IS the climax. Photorealistic, 24fps, wide anamorphic, matching the color grade and atmosphere of Shot 02.
ATTACH: Car ref (V2 side profile) + any successful Shot 02 output to match the environment
Shot 14
The Flicker — Kira's Eyes
Post-Finish
Hero
Fade from black: Extreme close-up of Kira's eyes. No helmet. No visor. Dark brown eyes, perspiration on skin, scar above left eyebrow. Neutral lighting — no cyan, no red. She breathes. She blinks. For exactly 1–2 frames, a faint cyan light flickers in her iris. Could be a track light. Could be a reboot. Could be nothing. Eyes return to dark brown. She stares into camera. Hard cut to black.
ECU — Eyes Only
4 sec
Static
Sora 2 is strongest at subtle facial emotion and micro-expressions. Generate the base close-up with neutral warm lighting, exhaustion, and relief in her expression. The blink must look completely natural.
The 1–2 frame cyan reflection must be added in post. Do NOT generate it in the AI model — it needs to be precise enough that half the audience misses it on first viewing. Composite a faint cyan light catch in one iris only. At 24fps, 2 frames = 0.08 seconds.
Video Generation Prompt — Sora 2 + Post Compositing
STEP 1 (Sora 2): Fade from black. Extreme close-up of a Latina woman's face, early 30s. No helmet, no visor. Only her eyes, nose bridge, and eyebrow area fill the frame. Dark brown eyes, warm neutral lighting, perspiration on skin. She breathes slowly: one inhale, one exhale. She blinks naturally, once. Her expression holds exhaustion, relief, and the faintest edge of something unresolved. Static camera. No movement except her breathing and the blink. 4 seconds. Warm amber light from above, no cyan, no red, no digital elements. Photorealistic, anamorphic, extreme macro detail on eyes.
STEP 2 (Post-Production): Take the rendered video. Identify the blink frame. On the 1-2 frames immediately after her eyes reopen, composite a faint cyan light reflection in the iris of her LEFT eye only. The reflection should look like a point of light catching in the moisture of her eye, not a glow or overlay. At 24fps, 2 frames = 0.08 seconds. Test at playback speed. If you notice it easily, it's too strong. Reduce opacity until it's barely perceptible. The goal: half the audience misses it entirely on first viewing.
ATTACH: Kira Soul ID + Prompt 16 key frame (The Flicker still, warm neutral lighting, no digital elements)
Shot 15
Title Card — THE LAST DRIVER / PARITY
Design Element
Supporting
Hard cut to black. Text fades in: "THE LAST DRIVER" then below it: "PARITY." Clean, minimal. The Higgsfield watermark visible. After the flicker, these words carry a second meaning.
Title Card
7 sec
Create as a designed graphic element. Use Bebas Neue or similar bold condensed font. Animate in Cinema Studio or external editor. Keep it clean — the film's power does the work.
🎯 Phase 1 Test Generation Order
Start with the shots that will tell you the most about what Higgsfield can deliver. If these three work, the rest will follow.
1
Shot 12: The Silent Drift (The Fold)
The money shot. If Veo can deliver the dark beauty of a car drifting through rain with no artificial light, you know the film will be visually stunning. This is the poster frame.
Veo 3.1
2
Shot 02: The Futuristic Grid at Ironwood Raceway
The world-building shot. If Veo can create a convincing futuristic grid on a night track with holograms and wet asphalt, the entire world is established. This proves the concept.
Veo 3.1
3
Shot 11: "No." (The Plunge)
The performance shot. If Sora/Kling can deliver the facial emotion + lighting transition (red to black), the emotional climax will land. Test both models here.
Sora 2 + Kling 3.0
4
Shot 05: The Spine
The speed showcase. Tests whether AI can generate convincing high-speed racing with elevation changes. This determines how much pure racing action the film can include.
Veo 3.1 + Kling 3.0
5
Shot 01: Kira's Eyes
The character anchor. Lock the look of Kira — once you have her face, you use it as reference for every other shot. Soul for still → Sora for animation.
Soul / NBP → Sora 2
Post-Production Continuity Pipeline
Once individual shots are generated, use these tools to ensure consistency, apply VFX, and assemble the final edit:
A
Character Lock + VFX Editing
Kling O1
Element Library locks Kira's face across all shots. Semantic editing to swap backgrounds, add rain, remove inconsistencies. Motion transfer between shots.
B
Shot Assembly + Transitions
Cinema Studio
Timeline editing, start/end frame transitions between shots, multi-shot narrative logic. The editing backbone.
C
Precision Object Placement
Banana Placement
Place AEGIS logos, race number graphics, or sponsor elements into specific frame positions using draw-to-video storyboards.
D
Audio Integration
Wan 2.5
Native audio generation for dialogue ("No."), engine sounds, rain ambience, and the critical silence moment. Lip-sync for hero shots.
PARITY — Phase 1 Production Guide — Built for Higgsfield Contest