连续剧agent
第二步：
安装剧本资产skill
把下面的安装包skill发给codex，并附上提示词：
This content is only supported in a Feishu Docs
请安装 story-asset-builder 规则库
操作要求：
1. 找到我提供的 story-asset-builder.zip。
2. 解压到当前 Codex 工作区，最终路径应为：
./story-asset-builder/SKILL.md
3. 确认目录结构存在：
- ./story-asset-builder/SKILL.md
- ./story-asset-builder/templates/
- ./story-asset-builder/profiles/
- ./story-asset-builder/rules/
- ./story-asset-builder/examples/
- ./story-asset-builder/quick_prompts/
4. 读取 ./story-asset-builder/SKILL.md，理解它是“剧本资产拆解 + 提示词编译 + LibTV 画布规划”的规则库。
5. 以后当我说“使用 story-asset-builder”时，你要：
- 先解析剧本
- 拆解人物、场景、道具、怪兽/生物、载具、色卡、分镜资产
- 根据 profiles 选择项目风格
- 根据 templates 生成标准提示词
- 根据 rules 规划 LibTV 画布节点
- 最后调用当前工作区已有的 LibTV CLI 创建画布节点
6. 不要重新安装 LibTV CLI。
7. 不要读取、打印、修改我的 LibTV 密钥。
8. 不要把密钥写进任何文件。
9. 如果需要调用 LibTV，只使用当前已经测试可用的 CLI：
libtv-windows-amd64\libtv-win-x64\libtv.exe
完成后告诉我 story-asset-builder 是否安装成功，以及能否读取 SKILL.md。
第三步：
把剧本和提示词发给codex，开始自动生成
[Image]
使用 story-asset-builder。
项目 Profile：default_cinematic。
根据下面短剧创意创建 LibTV 资产准备画布：
一个男人醒来后发现自己被困在地下储藏室，门外有人敲门，但声音越来越不像人。
要求：
1. 自动拆解资产
2. 每个资产生成标准 Prompt
3. 调用当前工作区已有的 LibTV CLI 创建画布节点
4. 所有节点必须放到 LibTV 画布上
5. 先不要生成最终视频，只准备资产图节点
万物生原创提示词参考：
1.人物设定图提示词
[Image]
【prompt】
A character design reference sheet. Vertical 9:16 aspect ratio. Maximum resolution output. Pure solid neutral light-warm-grey studio backdrop. Single-sided side-rim backlighting from camera-right at 45 degrees behind the subject, no fill light, no front light, no other light sources—dramatic cinematic studio lookbook lighting.
【Layout】
2x2 grid with unequal panel heights:
- Top row: two smaller panels (each approximately 1/3 of total canvas height) — head/face close-ups
- Bottom row: two larger panels (each approximately 2/3 of total canvas height) — full body shots
- Thin neutral light-grey dividing lines between all four panels
- All four panels share the same solid backdrop and same lighting setup
【4 Panels】
Top-left panel — FRONT FACE CLOSE-UP:
Tight close-up of the female lieutenant's face, framed from approximately collarbone-line up to top of head. Subject facing camera directly. Side-rim backlight catches the right edge of the face/hair (camera's left side of frame), leaving the left side of the face in deeper shadow with subtle ambient bounce. Both eyes intact and visible, sharp piercing veteran gaze toward camera. Face fully visible—no headwrap covering nose or mouth, headcloth pulled down loose around shoulders/neck.
Top-right panel — SIDE FACE CLOSE-UP (PROFILE):
Tight close-up of the lieutenant's face in pure side profile, framed from collarbone-line up to top of head. Subject facing camera-right (so the right side of her face faces camera-right and the LEFT side of her face faces camera). Side-rim backlight catches the back of the head and silhouettes the profile against the backdrop. Long black braided hair visible falling down the back/side, tribal hair ornaments (small bones, beads, metal fragments woven in) catch the rim light. Face fully visible.
Bottom-left panel — FULL BODY FROM NECK DOWN, FRONT VIEW (NO FACE):
Full body shot from collarbone-line down to feet, facing camera directly. The face/head is intentionally cropped out of frame at the top of this panel—only the neck, shoulders, torso, arms, hips, legs, and feet are visible. This panel is dedicated to body proportions, costume, equipment, weapons, and the mechanical LEFT shin/calf prosthetic—zero face involvement so this panel survives any future face replacement edit. Side-rim backlight catches the right shoulder and right side of body, defining athletic lean silhouette. Dual curved sand-blades held one in each hand or sheathed at hips, all front-side accessories visible.
Bottom-right panel — FULL BODY BACK VIEW (INCLUDING BACK OF HEAD):
Full body shot from head to feet, facing directly away from camera (back of subject toward camera). Includes back of head, headwrap wrapping around skull, long black braid falling down the back to mid-back length. Short composite recurve bow + small arrow quiver strapped diagonally across back fully visible. Trophy decorations on back of belt visible (more refined and woven than the chieftain's—small braided strands, beads, polished bone fragments). Side-rim backlight catches right shoulder silhouette from this back angle.
【Subject — same character across all 4 panels】
A female desert raider lieutenant—the chieftain's second-in-command and most lethal close-combat warrior of the Sand Hunter tribe. Same tribal aesthetic family as the chieftain (weathered war-torn ancient lost-tech) but differentiated for individual identity.
Character Core:
Approximately 30-40 years old, seasoned combat veteran. Lean wiry athletic combat physique—sinewy and agile, NOT bulky muscular, NOT slender model-fashion thin, NOT voluptuous. Built like an experienced desert raider who survives by speed and precision.
ETHNICITY — CRITICAL: Central Asian / Turkic / Sogdian Silk Road ethnic features—the historically accurate "杂胡" mixed-tribe peoples of the Sui-Tang frontier desert (突厥+粟特+西域本土+汉流亡者混合). Specific features:
- Wide prominent cheekbones (Central Asian Turkic structure)
- Deep-set almond-shaped eyes with monolid or low-fold (NOT round Western double-fold eyes)
- Black or very dark brown eyes (NOT blue, NOT green, NOT grey, NOT hazel)
- Black hair straight or slightly wavy texture (NOT blonde, NOT red, NOT light brown)
- Weathered tan-bronze skin from desert sun (warm bronze undertone, NOT pale Northern European fair, NOT dark African)
- Strong jawline with full but defined cheek-cheekbone structure (NOT delicate Western pointed chin, NOT soft round modern East Asian)
- Straight or slightly aquiline nose with moderate bridge (NOT thin pointed Western Caucasian nose, NOT flat-bridged modern East Asian nose—more of a Sogdian merchant Silk Road profile)
- Thick natural eyebrows, weathered skin texture with sun lines
Battle scars on face: small diagonal scar through right eyebrow + a thin healed scar on left jawline (these add veteran toughness without disfiguring the features—both eyes fully intact and functional).
Hair: long black hair gathered into a single thick braid down the back to mid-back length, tribal hair ornaments woven in (small bones, polished beads, small metal rings, small leather wraps). Hair partially tied back to keep face clear, with a few loose strands at temples.
Headwear: tribal headwrap of weathered tan-and-rust desert cloth, partially wrapped around the head and back of neck but FACE FULLY UNCOVERED in face close-up panels. Wrap can extend down as a hood/scarf around shoulders.
Upper body: weathered dark leather sleeveless vest cut short above the navel (functional desert combat cut, NOT decorative), exposing toned midriff and arms with battle scars and faded tribal tattoo markings on shoulders, upper arms, and ribs.
Lower body: form-fitting weathered dark leather trousers or cropped breeches with knee-high desert boots, multiple belt straps and small pouches around waist (more numerous and smaller than the chieftain's—she carries more variety of small tools and ammunition).
Left shin/calf — local mechanical modification: the LEFT shin and lower leg from below the knee to the ankle replaced with a CRUDE PRIMITIVE BRONZE MECHANICAL PROSTHETIC—same patina aesthetic as the chieftain's right forearm and the wider Sand Hunter tribe's combat exoskeletons. War-torn brass-and-leather construction with visible bolts, leather strap bindings to upper leg stump, exposed gear mechanisms at the knee and ankle, slight verdigris patina (Mecha Crimson Bronze #7A6845, Aged Gold #846035 accent). The mechanical foot ends in a clawed/toed combat-boot-like structure. NOT precision cybernetic, NOT modern, NOT clean—desert blacksmith salvage repair aesthetic.
Trophy decorations: more refined than the chieftain's brutalist trophies—she wears woven strands of small polished bones, colored thread bindings, small beads, fragments of defeated enemies' jewelry/buckles. Trophies on belt, shoulder strap, and braided into hair.
Footwear: rugged desert leather boot on right foot wrapped with cloth bindings around the calf (matching the chieftain's style); the left foot is the bronze mechanical prosthetic itself.
【Weapons - visible in body panels】
- Primary: dual curved scimitar-style sand-blades—two matched single-handed curved blades, sand-pitted iron, bone-and-leather wrapped handles, designed for fast dual-wielding close combat. One held in each hand or sheathed crossed at the back hips. Distinctly NOT identical to the chieftain's massive heavy two-handed war blade—hers are agile, fast, paired.
- Secondary: belt of 5-6 small dart-style throwing blades or poisoned darts (smaller and more numerous than the chieftain's throwing knives), small pouch of additional projectiles
- Ranged: short composite recurve bow (smaller than the chieftain's full-size bow, designed for quick draw) strapped diagonally across back, small arrow quiver attached
【Color Palette】
Desert sand tan, deep burnt umber leather, aged bronze patina (Mecha Crimson Bronze #7A6845 + Aged Gold #846035), dried-blood-red tribal accent on cloth bindings, accent of dark turquoise stone bead detailing in jewelry/hair (Silk Road trade material). ALL materials weathered war-torn dust-coated, NOT clean NOT polished.
【Pose - same across all 4 panels】
Default standing pose, agile ready stance—weight balanced on both feet, body slightly turned in athletic readiness, hands resting on or near weapon hilts at hips. Subtle predatory alertness/calm authority—NOT aggressive action pose, NOT relaxed civilian, NOT seductive/sensual pose, NOT fashion-model pose. The body language reads as "an experienced lethal warrior at quiet rest."
【Critical consistency】
SAME individual character in all 4 panels—identical face structure (Central Asian/Turkic/Sogdian features), identical scars (right eyebrow + left jawline), identical hair (single thick black braid + tribal ornaments), identical clothing items in identical positions, identical weapons in identical mounts, identical mechanical LEFT shin prosthetic, identical trophy decorations. Same studio backdrop, same single-sided side-rim backlighting setup across all 4 panels.
【Lighting - critical】
Single-sided side-rim backlighting from camera-right at 45-degrees behind subject. NO fill light. NO front light. NO ambient. NO other light sources. Strong dramatic edge-light catches the right side of subject silhouette in every panel, leaving the left side of subject in deep falloff shadow with subtle ambient bounce. Cinematic high-end fashion-lookbook studio photography (Peter Lindbergh / Annie Leibovitz portrait style with single rim source). Background remains evenly lit solid neutral light-warm-grey backdrop.
【Style】
Hyperrealistic photographic character design rendering—shot like a high-end practical costume photographed in a film production studio for a cinematic lookbook reference board. NOT illustration, NOT obvious 3D CGI render.
Reference aesthetics for Central Asian Silk Road ethnic accuracy:
- 《长安十二时辰》(The Longest Day in Chang'an, 2019) 西域狼卫 — Sui-Tang Silk Road frontier desert peoples, the most accurate ethnographic anchor for this character's facial features and tribal aesthetic
- Mongol: The Rise of Genghis Khan (2007) — Central Asian battle-hardened warrior physique and weathered features (Khulan Chuluun's portrayal of the warrior queen Khulan as facial reference)
- Monster Hunter Wilds desert hunter aesthetic — for the tribal armor/weathered-leather/bone-trophy DNA
Light Kodak Vision3 35mm film grain texture overlay.
【Quality】
Maximum resolution output. 8k ultra clean rendering. Sharp clean edges. Every leather strap, every scar, every bone trophy, every mechanical bolt on the prosthetic clearly defined. Real physical costume cinematic photography aesthetic.
【Panel Labels】
Small clean monospace sans-serif black text below each panel:
- Top-left: "FRONT FACE"
- Top-right: "SIDE FACE"
- Bottom-left: "BODY FRONT (NO FACE)"
- Bottom-right: "BACK"
Title at top center of canvas (small clean black sans-serif): "SAND HUNTER LIEUTENANT — DEFAULT FORM CHARACTER SHEET"
【Constraints】
Ethnic accuracy (CRITICAL): NOT pure Caucasian European face, NOT blonde or red or light brown hair, NOT blue or green or grey or hazel eyes, NOT pale fair Northern European skin, NOT thin pointed delicate Western nose, NOT round Western double-fold eyes, NOT pure Han Chinese face, NOT modern K-pop or Chinese drama actress aesthetic, NOT cute kawaii anime face. MUST be Central Asian Turkic/Sogdian Silk Road mixed ethnicity with appropriate weathered desert-warrior features.
Body & costume: NOT bulky muscular (lean athletic agile only), NOT slender model-fashion thin, NOT voluptuous sensualized body, NOT seductive pose or framing, NOT young teen-aged appearance (must be 30-40 weathered veteran).
Aesthetic: NOT precision cybernetic prosthetic (primitive crude war-torn bronze only), NOT medieval European armor, NOT modern military uniform, NOT Ming-Qing imperial costume, NOT samurai armor, NOT belly dancer or fantasy-game-female-warrior chainmail-bikini aesthetic, NOT clean polished anything, NOT shiny new chrome (weathered patina only), NOT cyberpunk neon, NOT obvious CGI 3D render, NOT cartoon illustrated, NOT anime style, NOT colorful saturated palette (muted desert only).
Layout & lighting: NOT 1:1 square aspect (must be vertical 9:16), NOT equal-sized 4 panels (top row smaller, bottom row larger), NOT face visible in bottom-left "BODY FRONT (NO FACE)" panel (face strictly cropped out at top of panel), NOT face covered by headwrap in face close-up panels (face fully visible), NOT colorful background (solid neutral light-warm-grey only), NOT soft even neutral lighting (must be single-side rim backlight only), NOT front light, NOT fill light, NOT any text other than the four panel labels and the title.
2.道具图提示词
[Image]
=== GPT-Image-2 Prompt - 万物生第1集 沙猎部落 机械赛加羚羊 兽形态设计图 v1 ===
【格式】
- 16:9横版+最高分辨率
- 4格均等2x2 = SIDE + FRONT + TOP + 3/4 VIEW
- 同studio backdrop+single-side rim backlight环境（继承沙猎部落photoshoot DNA）
【兽设定】
精英沙匪搭档——**机械赛加羚羊（Mechanical Saiga）**——独立形态。本张是兽自身的兽形态人设，合体外骨骼形态后续做。
- 种属：纯机械（古代失落科技遗物+无生物组织+半智能AI）
- 战术：群体冲撞+蹄踏+gallop颠簸
- signature视觉记忆点：奇特bulbous长鼻（机械化）+螺旋竖角
- 体型：肩高1.2米+体长1.5米（真实赛加放大约1.5倍——保留比例不变成马）
- DNA：和合体外骨骼+人物机改+沙艇全部同青铜patina
---
【prompt】
```
A creature design reference sheet. Horizontal 16:9 aspect ratio. Maximum resolution output. Pure solid neutral light-warm-grey studio backdrop. Single-sided side-rim backlighting from camera-right at 45 degrees behind the subject, no fill light, no front light, no other light sources—dramatic cinematic studio lookbook lighting. Same studio environment as the Sand Hunter character sheets and sand-skiff design sheet—reads as part of the same photoshoot session.
【Layout】
2x2 equal grid:
- Top-left: SIDE VIEW (full broadside profile, the primary view)
- Top-right: FRONT VIEW (head-on, showing facial features prominently—the bulbous nose and horns)
- Bottom-left: TOP-DOWN VIEW (god's-eye view from directly above showing footprint and dorsal mechanical layout)
- Bottom-right: 3/4 PERSPECTIVE VIEW (three-quarter angle showing depth and dimensionality, hero shot)
Thin neutral light-grey dividing lines between all four panels. Creature centered in each panel with even space margin.
【Subject — same creature across all 4 panels】
A mechanical Saiga antelope—an ancient lost-technology mechanical war-beast, modeled after the Central Asian Saiga antelope species, retrofitted as a war-mount for the Sand Hunter tribe's elite raiders. NOT a biological creature, NOT a cyborg with biological tissue—**fully mechanical construction with no organic components**, but with semi-sentient AI consciousness reactivated by the tribe (a relic of a long-lost mechanical civilization, now serving new masters).
SPECIES IDENTIFICATION (CRITICAL):
Saiga antelope (Saiga tatarica)—the Central Asian steppe-desert species native to the Sui-Tang frontier territories. Highly specific anatomy:
- Bulbous oversized proboscis-like nose hanging down past the mouth, soft and drooping in life (in this mechanical version: rendered as a bronze mechanical proboscis with visible air-filtration intake slits and small exhaust ports along its length—the signature feature of the species, the most critical recognition marker—if this nose is missing or replaced with a regular antelope muzzle, the design fails)
- Vertical spiraled horns on males—straight-up-pointing horns with translucent ringed spirals (in this mechanical version: rendered as bronze mechanical horns with the original spiraled ringed structure, sharpened at the tips for combat use)
- Compact stocky body with relatively short legs (NOT long-legged like a deer or gazelle, NOT slim like a normal antelope—the Saiga is built like a muscular sheep-antelope hybrid)
- NOT a regular antelope, NOT a deer, NOT a gazelle, NOT a goat, NOT a sheep—the Saiga has a unique unmistakable silhouette dominated by the bulbous nose
Scale: Approximately 1.2 meters at the shoulder, 1.5 meters body length nose-to-tail. Slightly larger than a real Saiga (about 1.5x scale) for war-mount viability, but maintaining authentic Saiga body proportions—NOT scaled up to horse-size (must read as Saiga, not horse-shaped antelope).
Construction (CRITICAL):
Fully mechanical, no biological tissue. The body is constructed of:
- Bronze mechanical bone-and-muscle armature: visible mechanical "skeleton" of bronze rods, joints, gears, hydraulic-style pistons (clockwork-driven not steam) where the biological skeleton would be—giving the creature its anatomical proportions
- Bronze panel armor "fur": the surface is covered in overlapping bronze panel-plates that approximate the silhouette of fur—not actual fur, NOT biological skin—segmented bronze panels with weathered patina, joint-line gaps showing mechanical movement underneath
- Mecha Crimson Bronze #7A6845 + Aged Gold #846035 patina with verdigris in seam recesses—exactly matching the Sand Hunter tribe's combat exoskeletons, prosthetics, sphere-companion, and skiff drive units (consistent universe DNA)
- Tribal geometric runic glyphs etched into the larger bronze panels (matching the Qingtong Sphere and tribe equipment glyph aesthetic—abstract pre-civilization mechanical-runic geometry, NOT Chinese decorative motifs, NOT real-culture runes)
Bulbous mechanical nose (CRITICAL signature feature):
A large pendulous bronze proboscis hanging down past the mouth line, approximately 25cm long, reproducing the iconic Saiga nose silhouette in mechanical form:
- Made of segmented bronze rings tapering toward the tip
- Visible slatted air-intake vents along the upper surface (the Saiga's biological function was filtering desert dust—the mechanical version retains this purpose with mechanical filters)
- Small exhaust ports at the rear sides emitting faint visible heat-shimmer or steam-trace
- Slight downward droop matching biological Saiga posture
- Verdigris patina in deep recesses
Horns:
A pair of vertical spiraled mechanical horns mounted on the head, approximately 30cm long, pointing nearly straight up with slight forward angle. Each horn is bronze with the characteristic Saiga ringed-spiral pattern preserved (the ridged twist visible). Sharpened metal tips for combat use. Slight battle-wear visible on the spiral edges.
Eyes:
Small round mechanical eyes with glowing cyan inner light #4FC3D9 (matching the Qingtong Sphere companion's eye glow—a deliberate visual hint at shared ancient lost-technology origin across all the universe's mechanical creatures). Bronze rim around each eye with thin glass-like lens cover. Eyes appear alert, semi-sentient, watching.
Legs and hooves:
Four mechanical legs with stocky proportions matching biological Saiga (relatively short legs for the body—NOT long-legged like a horse or deer). Each leg has visible bronze joint pivots at hip, knee, and ankle. Cloven mechanical hooves at the feet—two-toed bronze hoof shoes designed to grip sand. Visible wear and sand-pitting on the hoof surfaces. Lower legs show fine mechanical detail of the gallop-driving pistons.
Tail:
Short stubby mechanical tail (about 10cm)—matching biological Saiga's small tail. Bronze segmented construction.
Cockpit/合体接口 (CRITICAL — visible in TOP and 3/4 panels):
On the dorsal surface (back) between the shoulder blades, a clearly visible mechanical cockpit hatch / pilot integration interface—a roughly human-torso-sized rectangular-oval seam with bronze locking clasps around its perimeter, visible hinges along one edge, and small step-rungs on the side flank for mounting access. This is where the elite raider partner climbs in and integrates with the war-beast. The hatch is currently CLOSED in this character sheet (showing the design but in dormant/standalone configuration). Below the hatch line, faintly visible internal mechanical structure suggests the harness/cradle inside.
Tribal trophy decorations and partner-binding tokens:
Moderate quantity of trophy decorations affixed to the war-beast:
- Leather-strap harness around the neck/chest with woven beads, polished bone fragments
- A small leather pouch tied to the saddle-area near the cockpit hatch (containing the partner-raider's personal token—a small carved bone amulet visible peeking out)
- 3-5 small trophies (defeated enemies' equipment fragments) tied to flanks via leather cords
- Tribal pattern dyed-cloth banding around the lower legs and around the base of the horns
- Mecha Crimson Bronze + Aged Gold + dried-blood-red color accents
Weathering:
War-torn weathered—heavy verdigris patina in deep recesses, sand-pitting on bronze surfaces, small dents from past combat, mismatched repair patches in slightly different bronze tones (the tribe has maintained this ancient relic for many generations using whatever bronze they could find). NOT clean, NOT polished, NOT factory-new.
【4 Panels detail】
Top-left panel — SIDE VIEW: Pure broadside profile view. Beast facing camera-right (head to the right). Full body visible nose-tail. The bulbous mechanical proboscis prominently visible drooping down from the head. Vertical spiraled horns visible. Stocky leg proportions visible. Cockpit hatch on the back visible from this angle. Dorsal trophy decorations visible. Side-rim backlight catches the upper back, horn tops, and rear flank.
Top-right panel — FRONT VIEW: Beast head facing camera directly. The bulbous mechanical proboscis nose is the dominant feature in this panel—centered, drooping, with air-intake vents and exhaust ports visible. Horns visible rising vertically with slight forward angle. Cyan eyes glowing on either side. Front legs visible flanking the chest, cloven hooves on the ground.
Bottom-left panel — TOP-DOWN VIEW: God's-eye view from directly overhead looking down at the beast. Shows the full footprint and dorsal mechanical layout. The cockpit hatch on the back between the shoulder blades is most clearly visible in this view—the rectangular-oval seam with locking clasps. Tribal pattern bronze panels visible in their full layout. Horns visible from above as paired vertical spires. Stocky body proportions clearly readable.
Bottom-right panel — 3/4 PERSPECTIVE VIEW: Three-quarter angle showing depth, approximately 30-45 degrees from front-side, slightly elevated viewpoint. Side profile + front + slight top all visible. The signature bulbous nose visible at this angle. Dorsal cockpit hatch visible from the elevated angle. Cyan eye glow catches the rim light. The "hero shot" of the four views—dynamic stance with one foreleg slightly raised as if mid-stride.
【Critical consistency】
SAME individual creature in all 4 panels—identical body proportions (stocky Saiga build), identical bulbous proboscis, identical horns, identical cockpit hatch position and design, identical trophy decoration arrangement, identical bronze panel pattern, identical weathering pattern. Same studio backdrop, same single-sided side-rim backlighting setup across all 4 panels.
【Lighting - critical】
Single-sided side-rim backlighting from camera-right at 45-degrees behind subject. NO fill light. NO front light. NO ambient. NO other light sources. Strong dramatic edge-light catches the right side of the creature silhouette in every panel, leaving the left side in deep falloff shadow with subtle ambient bounce. Cyan eye glow self-illuminates faintly on bronze panels adjacent to the eyes.
【Style】
Hyperrealistic photographic creature design rendering—shot like a high-end practical animatronic creature photographed in studio for a film production design board. NOT illustration, NOT obvious 3D CGI render.
Reference aesthetics:
- Real-world Saiga antelope wildlife photography (for accurate species anatomy—the bulbous nose, vertical spiral horns, stocky proportions are mandatory)
- Studio Ghibli Princess Mononoke spirit-beast aesthetic (for the ancient mythical-mechanical creature feel)
- Ancient bronze Chinese ceremonial vessels and Han dynasty tomb mechanical figures (古代青铜机关——for the lost-tech bronze patina and ringed-glyph aesthetic)
- Pacific Rim mecha weathered combat aesthetic (for the war-torn salvage maintained-by-primitive-hands feel)
- Monster Hunter Wilds / Horizon Zero Dawn mechanical desert beasts (for the contemporary mechanical-fauna aesthetic reference)
Light Kodak Vision3 35mm film grain texture overlay.
【Quality】
Maximum resolution output. 8k ultra clean rendering. Sharp clean edges. Every bronze panel seam, every gear tooth on visible joints, every air-vent slat on the proboscis, every spiral ring on the horns, every leather strap on the trophy decorations clearly defined. Real practical-creature animatronic photography aesthetic.
【Panel Labels】
Small clean monospace sans-serif black text below each panel:
- Top-left: "SIDE"
- Top-right: "FRONT"
- Bottom-left: "TOP"
- Bottom-right: "3/4 VIEW"
Title at top center of canvas (small clean black sans-serif): "MECHANICAL SAIGA — WAR-MOUNT DESIGN SHEET"
【Constraints】
Species accuracy (MOST CRITICAL — high failure risk):
MUST be a Saiga antelope—the bulbous proboscis-like nose is the absolute mandatory recognition feature. NOT a regular antelope, NOT a deer, NOT a gazelle, NOT a goat, NOT a sheep, NOT an ibex, NOT a horse, NOT a cow. Vertical spiraled ringed horns (NOT curved horns, NOT spiral-curved like a kudu, NOT branched like deer antlers). Compact stocky body proportions (NOT long-legged like a horse or deer).
Mechanical purity:
Fully mechanical construction with NO biological tissue—NOT a cyborg, NOT biological-with-armor, NOT fur-and-flesh-with-mechanical-parts. ALL surfaces must be mechanical bronze-and-metal panels and components. NO real fur, NO real skin, NO real muscle, NO real eyes (mechanical eyes with cyan glow only).
Aesthetic anchoring:
NOT precision modern robotics, NOT clean factory-new metal (war-torn weathered patina only), NOT shiny chrome (aged bronze with verdigris only), NOT steampunk Victorian (this is ancient lost-tech NOT industrial-revolution), NOT cyberpunk neon, NOT obvious CGI 3D render, NOT cartoon, NOT anime, NOT fantasy elf-craft elegance, NOT pure militaristic mecha (must retain organic creature silhouette underneath the mechanical surface).
Cockpit hatch presence:
The dorsal cockpit hatch between shoulder blades MUST be visible in TOP-DOWN and 3/4 panels (this is the key story-element of the war-mount system). NOT a fully sealed creature—the partner-integration interface must be readable.
Layout and lighting:
NOT 1:1 square aspect (must be horizontal 16:9), NOT 9:16 vertical, NOT colorful background (solid neutral light-warm-grey only), NOT soft even neutral lighting (must be single-side rim backlight only), NOT front light, NOT fill light, NOT colorful saturated palette (muted desert bronze tones only), NOT any text other than the four panel labels and the title.
```
=== END ===
【实操注意】
1. 跑出来重点检查（按风险优先级）：
- 奇特bulbous nose（MAX风险）——Saiga的灵魂特征，AI模型很可能跑成普通羚羊muzzle。如果没跑出明显下垂的proboscis-like鼻，整图失败重做+按"the bulbous proboscis is the mandatory recognition feature, MUST be visibly oversized and pendulous, NOT a normal antelope muzzle"加强
- vertical spiraled竖直螺旋角——容易跑成弯曲羚羊角（如kudu）或鹿角分叉。必须是vertical+spiraled+ringed
- stocky compact体型——容易跑成长腿马或苗条鹿
- 纯机械（无生物组织）——容易跑成"fur-with-armor"生物机改。必须all bronze panels
- 背部cockpit hatch——这是和合体外骨骼系统的视觉钩子+必须在TOP/3-4 panels可见
- 同沙猎部落青铜DNA（#7A6845/#846035 patina+verdigris+部族纹）
- 眼睛青色发光#4FC3D9——同青铜球体小宠物的同源科技钩子
- 16:9横版+4格均等+SIDE/FRONT/TOP/3-4 VIEW
2. 跑歪兜底加强：
- 跑成普通antelope/deer/gazelle → "Saiga tatarica species, distinguished by oversized bulbous proboscis-like nose hanging past mouth (THIS IS MANDATORY), vertical spiral-ringed horns straight up, stocky compact body, NOT a regular antelope/deer/gazelle"
- 跑成生物机改（fur+armor）→ "fully mechanical construction, all bronze panel armor surface in place of fur, NO biological tissue, NO real fur, NO real skin"
- 跑成现代精良机器 → "ancient lost-technology bronze relic, weathered war-torn patina, salvage-maintained by primitive tribal hands, NOT modern robotics"
- 鼻子跑歪了 → "the bulbous mechanical proboscis is approximately 25cm long, hangs visibly past the mouth line in a soft droop, with slatted air-intake vents along upper surface and small exhaust ports at rear sides"
- 角跑成弯曲 → "vertical straight-up-pointing horns with characteristic spiraled-ringed surface (like the biological Saiga horns), NOT curved horns like kudu or impala, NOT branched antlers like deer"
- 看不到cockpit hatch → "dorsal mechanical cockpit hatch between shoulder blades clearly visible in TOP-DOWN and 3/4 perspective views, rectangular-oval shape with bronze locking clasps and hinges"
3. 关于"兽形态"vs"合体外骨骼形态"：
- 这张prompt是兽独立形态——cockpit hatch CLOSED+在dormant/standalone configuration
- 合体外骨骼形态（沙匪钻进去之后）后续单独做+基于这张兽形态做参考
3.怪兽对比图
[Image]
盲掘巨足类 · 多视图设定图 · 含人形比例对照
GPT-Image-2 生图 Prompt v4 (Character Sheet / Turnaround)
================================
Scene:
A professional creature concept design sheet on a clean, neutral light-grey background. Organized grid layout in the style of official VFX production design reference sheets used in films like Pacific Rim and Dune. Even, soft, consistent lighting across all views with no harsh shadows — orthographic concept art lighting designed to reveal subject form clearly.
Subject:
A colossal terrestrial isopod-like creature, approximately 35 meters long and 8 meters tall at the highest point of its arched back. The body consists of nine heavily armored dorsal segments forming a continuous arched spine. The creature is presented in four separate views on a single sheet, plus a scale comparison strip and a color palette strip.
Important Details:
Layout structure of the sheet:
LEFT HALF (large panel): Full side-profile orthographic view of the creature, facing frame-left, occupying the left half of the sheet. To the right of the creature within this same panel, place a simple black stick-figure silhouette of a standing adult human (1.8 meters tall) for scale comparison — the human silhouette should appear tiny next to the creature, roughly 1/20th of the creature's height. Small label text below the human silhouette reads "HUMAN 1.8M". Small label text below the creature reads "BODY LENGTH 35M".
TOP RIGHT (medium panel): Top-down orthographic view of the creature seen from directly above, clearly showing the nine segmented carapace plates along the spine and the eighteen digging limbs (nine pairs) extending symmetrically from the body sides. Label text reads "TOP VIEW".
MIDDLE RIGHT (medium panel): Front-facing close-up view of the head, showing the massive forward-protruding keratin shovel-jaw like a thick iron spade with chipped and notched edges, the horizontal mouth opening beneath, and the complete absence of eyes. Label text reads "HEAD DETAIL — SHOVEL JAW".
BOTTOM STRIP (horizontal panel spanning the full width): A scale comparison row showing five black silhouettes lined up left to right with size labels — a human figure (1.8m), an african elephant (3m), a city bus (12m), a blue whale (30m), and the creature itself (35m). Each silhouette labeled with its name and height/length. The creature silhouette is the largest and matches the side-profile shape from the main panel.
BOTTOM RIGHT CORNER: A small horizontal color palette strip showing 5 swatches of the creature's surface colors, each labeled with a hex code: pale sand-yellow #C9A876, warm tan #A8855A, weathered brown #7A5D3F, dark earth-brown #4A3826, and shadow membrane grey-brown #2D2218.
Creature design specifications (applies to all views):
The dorsal carapace is coarse, porous, weathered stone-like in texture, sand-yellow to dark brown in coloration with pale wind-erosion striations across each segment. The 3rd and 6th carapace segments bear small parasitic crustacean growths and dried barnacle-like encrustations clustered along their ridges. Segment joints reveal darker softer-looking membrane tissue in the gaps between armored plates. Each of the nine body segments bears a pair of short, thick, powerful digging limbs ending in broad shovel-shaped digging claws. The head has no eyes; only the massive forward-protruding keratin shovel-jaw with a horizontal mouth opening beneath. The underside is paler and less armored with a row of horizontal vent-like openings along the abdomen.
Use Case:
Official creature design reference sheet for a science fiction film production. Intended as a single-page deliverable for the production design team and VFX vendor to model the creature accurately, with built-in scale references and color specifications. Must read clearly as a professional concept art production sheet.
Constraints:
Aspect ratio 16:9. Clean light-grey background (#E8E6E2) across the entire sheet. All view panels separated by thin neutral grey divider lines. All text labels rendered in clean sans-serif typography, dark grey color (#2A2A2A), small but legible — label text must be sharp and readable. Photorealistic creature rendering within each view panel, with 35mm film grain texture. Neutral color grading across all panels. Consistent lighting direction and intensity across all four creature views — no view should appear lit differently from the others. Sharp focus across all creature details. No environmental elements (no sand, no sky, no dust, no atmosphere) — all creature views float on the clean background. No additional creatures, no human figures other than the small black silhouettes specified for scale reference, no vegetation, no decorative elements beyond what is specified in the layout. Orthographic projection for all creature views — no perspective foreshortening.
4.视频生成影调色卡
[Image]
A clean digital color palette reference design, 16:9 horizontal layout, pure flat design aesthetic with no photographic elements.
The background is a solid neutral off-white #F5F2EC fill, completely flat, no texture, no shadow, no paper, no surface.
The composition contains exactly three elements:
ELEMENT 1 - Title at the top center:
Clean sans-serif typography in solid black #1A1A1A, the text reads:
"WANWUSHENG · EP2 DESERT SURVIVAL · DAYLIGHT PALETTE"
Below the main title, in smaller lighter gray #888888 sans-serif:
"Sun-baked Arrakis tonality · 13 color system"
ELEMENT 2 - A single horizontal row of 13 rectangular color swatches, evenly spaced, centered vertically in the frame:
Each swatch is a clean flat rectangle of solid color fill, identical dimensions, sharp clean edges, no gradient, no shadow, no border.
The 13 swatches from left to right with these exact HEX fills:
01: #F4E6C8
02: #D4B896
03: #A89472
04: #6B5A47
05: #E8E0D2
06: #EAE2D0
07: #2A2620
08: #7A2F2C
09: #D8C9A8
10: #4A332A
11: #A82A2A
12: #6F1F1F
13: #4FC3D9
ELEMENT 3 - Two lines of small text directly beneath each swatch:
First line is the HEX code in clean monospace typography, color #333333.
Second line is the function label in even smaller clean sans-serif typography, color #666666, describing the color's role.
The 13 function labels from left to right:
01: hot sunlight cast
02: sand highlight
03: sand midtone
04: hard shadow
05: bleached sky
06: dust-filmed white
07: heat-bleached black
08: faded burgundy
09: cream linen
10: cracked leather
11: cinnabar accent
12: dark blood
13: LED cyan (dormant)
OVERALL DESIGN SPECIFICATION:
This is a flat 2D digital graphic design layout, the kind produced in Figma or Adobe Illustrator for a film production design document. No depth, no perspective, no lighting, no photographic rendering, no paper texture, no physical object. Pure flat vector-style color swatches on a flat background with crisp typography. Think production design bible reference sheet, color management document, or design system documentation.
The aesthetic is precise, clinical, informational, minimal, professional graphic design — not artistic, not painterly, not photographic.
Avoid: photographic rendering, paper texture, desk surface, table surface, depth of field, lens blur, film grain, photorealistic, 3D render, lighting effects on swatches, shadows under swatches, glossy finish on swatches, hand-painted look, watercolor look, illustration, sketch, mockup of physical object, real-world object simulation.
视频生成提示词：
[Image]
[STYLE LOCK]
35mm Kodak Vision3 500T film stock胶片质感,Greig Fraser掌镜,Dune 2021户外白天美学,skip-bleach negative LUT,模拟胶片颗粒,anamorphic宽银幕镜头质感,真实摄影皮肤纹理。
部落崖壁市集横扫段落,横向轨道镜头lateral truck,2.39:1宽银幕画幅,主光源为午后偏侧的暖金色直射阳光,投下硬阴影。
音轨:铁匠锤击金属的清脆铛铛声、市集人群的嘈杂叫卖与交谈、远处人群欢呼歌唱声浪、铁匠摊位木架颤动和铃铛震响、人群快速走过时的脚步声。
@[IMG_GEN_20260521223005_51851989_1] 作为刘回视觉锚定。
@[8a3a9cc439b45b2a3d05933409154700] 作为整段画面色卡锚定+所有色彩严格按色卡24色执行+不可偏离色板+本段重点调用炉火映照#E87838到深炽红#A8401C的暖橙暖红主光+以及硬阴影暖灰#6B5A47到深棕#4A3E32的阴影层+人群衣物上的干血暗酒红#6B3A2E与沙猎赭#A8744A与烟熏炭灰#3A322A+以及朱砂红#A82A2A作为人群中点缀的视觉钉子。
@[IMG_GEN_20260521050623_8cbd5082_1] 作为部落崖壁环境视觉锚定。
@[IMG_GEN_20260521082656_d62e4d48_1] 作为部落人群服装风格视觉参考,人群中每一个个体的服装组合、皮革绑带、铃铛饰物、青铜挂件、颜料色块标记都从这张物料图中随机抽取并组合,共享同一套部落服装语言,但人群中每一个人的长相、年龄、性别都各不相同,既有年轻男性也有年轻女性、既有中年壮汉也有中年妇人、既有白发老者也有半大少年,体型有高有矮有壮有瘦,肤色深浅各异,整体读作一个真实多样化的部落群体。
@[IMG_GEN_20260522002340_9be8f3fa_1] 作为铁匠视觉锚定。
画面从大面积尘奶白与沙猎赭与朱砂红混合的部落彩旗布料完全占据镜头开始,布料质感与上一段结尾的彩旗完全延续。在画面第一秒,这面彩旗被一阵气流向画面下方扯落,布料快速滑出画面底部,露出布料背后崖壁前的部落露天市集景象。镜头机位在两排摊位之间的通道中,以缓慢均匀的速度向画面左方做横向轨道平移lateral truck,机位高度大约在普通成年部落人的肩部高度。
镜头一开始向左移动,画面对面侧就立刻显露出本段的视觉主体:一位巨人铁匠和他的露天锻造摊。这位铁匠是整段镜头持续锁定的核心主体,他在画面中从第二秒开始就占据画面的视觉中心,并且在接下来的镜头横移过程中始终保持在画面构图的中央偏画面深处位置,直到第八秒前后镜头才逐渐离开他。
铁匠的体型异常巨大,他在画面中的纵向高度从画面最顶部一直延伸到画面最底部,完整占据整个画面的全部纵向高度,他的身体几乎是画面里所有其他人物高度的两倍多。紧贴在他摊位的左右两边各有一个正常体型的部落摊主,他们的整个身体从头到脚加起来只占据画面下方约三分之一的纵向区域,他们的头顶仅仅达到铁匠的腰部位置,他们站在铁匠身旁仰着头朝上看他,姿态像两个孩童站在大人身旁。铁匠摊位前方还有三到四个普通体型的部落人正在围观他锻造,全部仰头朝上看,他们之中最高的一位的头顶也只到铁匠肚腩的位置。铁匠本人呈现脂包肌特征——巨大的圆形大肚腩、宽阔得不成比例的胸膛、粗壮如柱的手臂、像盾牌一样宽的肩膀,他的整体轮廓呈现一个庞大圆润的人形,像一座小山一样矗立在普通人之中。
铁匠的一只手握着一把超规格巨锤,在镜头横移的过程中他连续重复挥动手臂锤击,锤头沉重地砸向砧上烧红的铁块,火星向四面八方迸溅。每一次锤落整个铁匠摊的木架都在剧烈颤动,挂在摊位上的小铃铛和金属物件被震得叮当作响,周围围观和邻摊的小人们的衣服和头发都被震波微微吹动。镜头在他面前缓慢横向滑过,从不同的横向位置展示他的庞大体型与周围正常人的对比。
在第七到第八秒之间,抬着刘回的部落人群以远快于镜头的速度从画面右侧涌入,从右向左快速涌过画面,经过铁匠摊位前方。人群中央那群健壮的部落人正高高抬着横躺在他们手掌上的刘回,刘回手中举着陶土酒壶继续大笑,即使是被人群高举到比正常人还要高的刘回的高度,也仅仅到达铁匠的胸膛位置,刘回看起来像被一群蚂蚁抬着经过一个巨人脚边。人群一闪而过,继续向画面左方涌去。
在第九秒到第十秒之间,一个特别近的部落人的宽厚背影从画面右侧近端入画,完全占据镜头前景,他穿着干血暗酒红色的厚重披风,披风上挂满铃铛饰物,他向左走的同时其背部布料快速扩张占据整个画面,在第十秒他的背部完全填满并遮挡整个镜头,画面终结于干血暗酒红色的布料色遮罩。
整段镜头中持续伴随铁匠锤头砸向砧上的清脆铛铛声clang clang clang、每次锤落带来的木架颤动声和铃铛被震得叮当响声、市集人群的嘈杂交谈与叫卖、远景人群的欢呼歌唱声浪、人群经过时的密集脚步声、铁匠摊位炉火燃烧的低沉呼呼声。