视频模板 - 电影叙事
game-cinematic, arpg, three-kingdoms, ancient-china, combat, cavalry, guanyu, key-visual, hud-safe, companion-to-image
提示词
A ~10 second in-engine cinematic ARPG action sequence, photoreal, Unreal-Engine-5-grade render quality, desaturated filmic color grading, shallow depth of field on the background. Strict in-game camera discipline — one continuous beat, no cuts, no swish pans.
# Scene (reference image: {argument name="reference_image" default="game-screenshot-three-kingdoms-guanyu-slaying-yanliang"})
- Setting: {argument name="environment" default="a broad open battlefield at dawn, dry yellow earth churned by thousands of hooves, low ground mist clinging to the field, distant spear forests on both sides, tattered war banners fluttering, a cold teal sky with warm amber sunrise light from the left"}.
- Hero (center-left of frame, stays identical from first to last frame): {argument name="hero" default="Guan Yu (关羽), a towering red-faced general in ornate green-lacquered lamellar armor with a long flowing beard, phoenix-eye expression, silk headband under a ridged helmet, red-green battle robe fluttering behind him"}.
- Mount: {argument name="mount" default="Red Hare (赤兔), a massive crimson-chestnut warhorse with a long mane, bronze-studded tack, full gallop"}.
- Weapon: {argument name="weapon" default="the Green Dragon Crescent Blade (青龙偃月刀), a long polearm with a heavy curved blade and a dragon-head socket, held in his right hand"}.
- Opposing general (mid-ground, center-right): {argument name="target" default="Yan Liang, an enemy general in dark iron armor on his own black warhorse, turning late to face the charge"}.
# Action (three micro-beats across ~10s, single continuous shot)
1. 0.0 - 4.0s Charge. Hero rides Red Hare full gallop directly toward camera-right; beard and robe stream backwards; hooves kick up dust trails; the polearm is held low and trailing so the crescent blade only catches a single glint of sunrise; camera tracks alongside on a smooth lateral dolly matching the horse's speed, framing the hero and horse in a clean center-left third.
2. 4.0 - 7.0s Raise + lock on. Hero rises slightly in the saddle and swings the Green Dragon Crescent Blade up into a shoulder-high guard; the blade catches a long sharp highlight; a faint {argument name="qi_color" default="cold jade-green"} qi glow builds along the edge; the enemy general is now fully in the mid-ground turning; their horses begin to cross paths; the background armies blur to a soft bokeh.
3. 7.0 - 10.0s Single clean cleave. Hero swings the blade in one decisive diagonal arc from high-right down to low-left as the two riders cross; a single bright gold-into-jade qi flash briefly blooms along the arc; the enemy general's helmet flies off-frame; his horse rears back; no blood, no gore, no on-screen text; hero's blade recovers smoothly to a trailing low guard; camera continues its lateral track and gently decelerates as the cleave finishes, ending on the hero's stoic profile riding past.
# Technical direction
- Framing: hero + Red Hare occupy the center-left third for the full shot; the enemy enters from the right third around the 4s mark; horizon low; never cut to a closeup.
- Camera: smooth lateral dolly matching horse speed, zero handheld shake, zero rack-focus; focus stays on the hero throughout.
- Lighting: cinematic cold-sunrise key from the left, warm amber rim on the armor, volumetric ground mist drifting left-to-right, subtle bloom on the qi glow only.
- Motion: photoreal horse gallop, 4-beat gait, plausible saddle bounce on the hero, blade weight feels real. The cleave is in real time — no slow-motion ramp. The qi flash lasts less than 0.2s.
- Color: desaturated filmic grade, cold teal + warm amber split, slight filmic grain, subtle chromatic aberration at frame edges only.
- HUD: none rendered in this shot (the companion image template carries the HUD; this video is a pure in-engine action beat).
# Negative prompt
no blood, no gore, no decapitated head shown on-screen, no severed limbs, no red splatter, no anime cel-shading, no comic speed lines, no slow-motion ram
常见问题
视频模板案例适合什么模型？
这条提示词主要面向 GPT Image 2，也可按效果迁移到同类图像模型。完整提示词可在 craftprompt.vip 复制。
使用「视频模板案例」需要参考图吗？
需要。请先准备风格或主体参考图，再粘贴提示词生成，效果更稳定。
如何用「视频模板案例」生成UI界面？
复制提示词 → 选择合适模型并上传参考图 → 在 craftprompt 生图工作台生成。