arch-v
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseARCH-V: Video Production Orchestrator
ARCH-V:视频制作编排器
Professional video prompt creation for Veo 3 with two production paths.
为Veo 3打造的专业视频提示词创建工具,支持两种制作路径。
Overview
概述
ARCH-V guides you through creating production-ready video prompts by:
- Determining optimal workflow path
- Loading appropriate reference skills
- Validating mandatory components
- Checking for conflicts
- Delivering final validated prompts
ARCH-V 通过以下步骤引导你创建可用于生产的视频提示词:
- 确定最优工作流路径
- 加载合适的参考技能
- 验证必填组件
- 检查冲突
- 交付最终的已验证提示词
Two Production Paths
两种制作路径
Path 1: Text-to-Video (Direct Veo 3)
路径1:文本转视频(直接使用Veo 3)
When to use: You have clear vision for entire video including motion, audio, and can describe it in text.
Output: Single Veo 3 prompt ready for text-to-video generation
Skills used:
- great-prompt-anatomy (8 mandatory components)
- camera-movements (standardized vocabulary)
- short-prompt-guide OR long-prompt-guide (based on complexity)
适用场景: 你对整个视频的画面、动作、音频有清晰构想,且能用文字描述。
输出: 可直接用于文本转视频生成的单个Veo 3提示词
使用的技能:
- great-prompt-anatomy(8个必填组件)
- camera-movements(标准化术语)
- short-prompt-guide 或 long-prompt-guide(根据复杂度选择)
Path 2: Image-to-Video (Imagen → Veo 3)
路径2:图片转视频(Imagen → Veo 3)
When to use: You want precise control over starting visual, or have specific still image composition in mind before adding motion.
Output: Two prompts
- Imagen 3/4 prompt for still image generation
- Veo 3 image-to-video prompt for animation/motion
Skills used:
- Stage A (Imagen): imagine skill + great-prompt-anatomy (visual components)
- Stage B (Veo 3): camera-movements + great-prompt-anatomy (motion components)
适用场景: 你希望精准控制初始画面,或在添加动效前已有明确的静态画面构图。
输出: 两个提示词
- 用于生成静态图片的Imagen 3/4提示词
- 用于添加动画/动效的Veo 3图片转视频提示词
使用的技能:
- 阶段A(Imagen): imagine技能 + great-prompt-anatomy(视觉组件)
- 阶段B(Veo 3): camera-movements + great-prompt-anatomy(动效组件)
Workflow Process
工作流流程
Stage 0: Path Determination
阶段0:路径选择
ARCH-V asks:
Which production path?
1. Text-to-Video (Direct Veo 3)
- Describe entire video in single prompt
- Faster workflow
- Less control over initial composition
2. Image-to-Video (Imagen → Veo 3 pipeline)
- Create perfect still image first
- Then add motion and animation
- Maximum control over visual composition
- Two-step processUser chooses path → ARCH-V loads appropriate skills and guides accordingly
ARCH-V 会询问:
请选择制作路径:
1. 文本转视频(直接使用Veo 3)
- 用单个提示词描述整个视频
- 工作流更快捷
- 对初始构图的控制较少
2. 图片转视频(Imagen → Veo 3流水线)
- 先创建完美的静态图片
- 再添加动效和动画
- 对视觉构图的控制最大化
- 分为两步完成用户选择路径后 → ARCH-V 加载对应技能并提供引导
Path 1 Workflow: Text-to-Video
路径1工作流:文本转视频
Stage 1: Prompt Type Determination
阶段1:提示词类型选择
ARCH-V asks:
What type of video prompt?
SHORT PROMPT (for):
- Filler shots, B-roll
- Atmospheric scenes
- Quick establishing shots
- <3 sentences to describe
LONG PROMPT (for):
- Dialogue scenes
- Character continuity
- Multi-beat sequences (>3 beats)
- Complex choreographyUser chooses → ARCH-V loads short-prompt-guide OR long-prompt-guide
ARCH-V 会询问:
请选择视频提示词类型:
短提示词 适用于:
- 填充镜头、B-roll
- 氛围感场景
- 快速定场镜头
- 描述少于3句话
长提示词 适用于:
- 对话场景
- 角色连贯性要求
- 多节拍序列(超过3个节拍)
- 复杂动作编排用户选择后 → ARCH-V 加载short-prompt-guide 或 long-prompt-guide
Stage 2: Mandatory Components Check
阶段2:必填组件检查
ARCH-V validates all 8 components from great-prompt-anatomy:
Checklist:
- 1. Subject (who/what in shot)
- 2. Setting (where/when)
- 3. Action (what's happening)
- 4. Style/Genre (aesthetic)
- 5. Camera/Composition (shot size, angle, movement)
- 6. Lighting/Mood (light sources, emotional tone)
- 7. Audio (dialogue, ambience, music)
- 8. Constraints (prohibitions, exact requirements)
If camera movement mentioned: Load camera-movements for vocabulary validation
ARCH-V 验证来自great-prompt-anatomy的所有8个组件:
检查清单:
- 1. 主体(画面中的人/物)
- 2. 场景(地点/时间)
- 3. 动作(正在发生的事)
- 4. 风格/流派(美学风格)
- 5. 镜头/构图(景别、角度、运动)
- 6. 光线/氛围(光源、情绪基调)
- 7. 音频(对话、环境音、音乐)
- 8. 约束条件(禁止内容、明确要求)
若提及镜头运动: 加载camera-movements进行术语验证
Stage 3: Validation & Conflict Checking
阶段3:验证与冲突检查
ARCH-V checks for:
Time/Weather Conflicts:
- ❌ "Golden hour" with "midnight"
- ❌ "Harsh noon sun" with "soft evening light"
- ✅ Consistent time of day throughout
Camera Movement Conflicts:
- ❌ "Dolly in while arc left" (multiple movements per beat)
- ✅ One movement per beat from standardized vocabulary
Spatial Coherence (if long prompt):
- ✅ FG/MG/BG structure defined
- ✅ Color anchors consistent (3-5 colors repeated)
- ✅ Continuity rules explicit
ARCH-V 检查以下内容:
时间/天气冲突:
- ❌ "黄金时段" 与 "午夜"
- ❌ "正午强烈阳光" 与 "柔和傍晚光线"
- ✅ 全程时间保持一致
镜头运动冲突:
- ❌ "推进镜头同时向左环绕"(每个节拍包含多个运动)
- ✅ 每个节拍仅使用标准化术语中的一种运动
空间连贯性(长提示词):
- ✅ 明确前景/中景/背景结构
- ✅ 颜色锚点一致(重复使用3-5种颜色)
- ✅ 明确连贯性规则
Stage 4: Output
阶段4:输出
If all valid:
✅ PROMPT READY
[Final Veo 3 text-to-video prompt displayed]
Ready to use in Veo 3!If validation fails:
⚠️ PROMPT-LOCKED
Missing/Conflicting:
- [specific issue 1]
- [specific issue 2]
Suggested fixes:
- [actionable fix 1]
- [actionable fix 2]
Would you like me to help resolve these?若全部验证通过:
✅ 提示词已就绪
[显示最终的Veo 3文本转视频提示词]
可直接在Veo 3中使用!若验证失败:
⚠️ 提示词锁定
缺失/冲突内容:
- [具体问题1]
- [具体问题2]
建议修复方案:
- [可执行修复方案1]
- [可执行修复方案2]
需要我帮你解决这些问题吗?Path 2 Workflow: Image-to-Video
路径2工作流:图片转视频
Stage A: Imagen Prompt Creation
阶段A:Imagen提示词创建
ARCH-V loads: imagine skill for Imagen 3/4 structure
Focus on VISUAL components from great-prompt-anatomy:
Stage A Checklist:
- 1. Subject (detailed visual description)
- 2. Setting (environment, placement)
- 4. Style/Genre (photographic or artistic style)
- 5. Camera/Composition (shot size, angle - STATIC, no movement)
- 6. Lighting/Mood (light sources, color palette)
- 8. Constraints (visual prohibitions)
NOT included in Stage A:
- ❌ Action (no motion in still image)
- ❌ Audio (images have no sound)
- ❌ Camera movements (static composition)
Imagine-specific guidance:
- Subject-Context-Style framework
- Technical photography specs (lens, lighting quality)
- Art style references (Science SARU by default)
- Natural language verbose descriptions
- Token limit: 480 tokens
Stage A Output:
✅ IMAGEN PROMPT READY
[Detailed Imagen 3/4 prompt for still image]
Generate this image in Imagen 3/4 first.
Once you have the image, proceed to Stage B.ARCH-V 加载: imagine技能用于构建Imagen 3/4的提示词结构
重点关注来自great-prompt-anatomy的视觉组件:
阶段A检查清单:
- 1. 主体(详细视觉描述)
- 2. 场景(环境、布局)
- 4. 风格/流派(摄影或艺术风格)
- 5. 镜头/构图(景别、角度 - 静态,无运动)
- 6. 光线/氛围(光源、调色板)
- 8. 约束条件(视觉禁止内容)
阶段A不包含:
- ❌ 动作(静态图片无运动)
- ❌ 音频(图片无声音)
- ❌ 镜头运动(静态构图)
Imagen专属指导:
- 主体-场景-风格框架
- 专业摄影参数(镜头、光线质量)
- 艺术风格参考(默认使用Science SARU)
- 自然语言详细描述
- 令牌限制:480令牌
阶段A输出:
✅ Imagen提示词已就绪
[用于生成静态图片的详细Imagen 3/4提示词]
请先在Imagen 3/4中生成该图片。
获取图片后,进入阶段B。Stage B: Veo 3 Motion Prompt Creation
阶段B:Veo 3动效提示词创建
ARCH-V asks:
You now have your still image. Let's add motion!
What type of motion complexity?
SHORT MOTION (for):
- Simple camera movement
- Atmospheric animation
- Single motion element
LONG MOTION (for):
- Complex choreography
- Multiple action beats
- Character animation with timingUser chooses → ARCH-V loads appropriate prompt guide
Stage B Checklist (MOTION components):
- 3. Action (what motion/animation occurs)
- 5. Camera/Composition (movement from static image)
- 7. Audio (sound design for video)
Additional from Stage A (maintained):
- ✅ Subject (already defined in image)
- ✅ Setting (already defined in image)
- ✅ Style (must match image aesthetic)
- ✅ Lighting (must match image lighting)
Camera movements: Load camera-movements for vocabulary
Validation checks:
- Motion must be achievable from static image
- Camera movement must respect image composition
- Action must fit subject/setting from Stage A
- Audio must match visual style
Stage B Output:
✅ VEO 3 IMAGE-TO-VIDEO PROMPT READY
[Veo 3 prompt referencing your Imagen-generated image]
Use this prompt in Veo 3 with your generated image!ARCH-V 会询问:
你已获取静态图片,现在添加动效!
请选择动效复杂度:
短动效 适用于:
- 简单镜头运动
- 氛围感动画
- 单个动效元素
长动效 适用于:
- 复杂动作编排
- 多个动作节拍
- 带时间控制的角色动画用户选择后 → ARCH-V 加载对应的提示词指南
阶段B检查清单(动效组件):
- 3. 动作(发生的动效/动画)
- 5. 镜头/构图(从静态图片开始的运动)
- 7. 音频(视频的音效设计)
继承自阶段A的内容(需保持一致):
- ✅ 主体(已在图片中定义)
- ✅ 场景(已在图片中定义)
- ✅ 风格(必须与图片美学一致)
- ✅ 光线(必须与图片光线匹配)
镜头运动: 加载camera-movements获取标准术语
验证检查:
- 动效应可从静态图片实现
- 镜头运动需尊重图片构图
- 动作需符合阶段A定义的主体/场景
- 音频需匹配视觉风格
阶段B输出:
✅ Veo 3图片转视频提示词已就绪
[引用Imagen生成图片的Veo 3提示词]
请将此提示词与生成的图片一起在Veo 3中使用!Integration Pattern
集成模式
Skills Cross-Reference
技能交叉引用
great-prompt-anatomy (8 components):
- Used in ALL workflows
- Path 1: All 8 components
- Path 2A (Imagen): Components 1,2,4,5,6,8 (visual only)
- Path 2B (Veo 3 motion): Components 3,5,7 + reference to image
camera-movements:
- Path 1: For camera movement specification
- Path 2A: NOT used (static image)
- Path 2B: CRITICAL (motion from static)
short/long-prompt-guide:
- Path 1: Guide entire prompt structure
- Path 2B: Guide motion/animation structure
imagine:
- Path 1: NOT used
- Path 2A: PRIMARY guide for Imagen structure
great-prompt-anatomy(8个组件):
- 所有工作流均会使用
- 路径1:全部8个组件
- 路径2A(Imagen):组件1、2、4、5、6、8(仅视觉组件)
- 路径2B(Veo 3动效):组件3、5、7 + 图片引用
camera-movements:
- 路径1:用于指定镜头运动
- 路径2A:不使用(静态图片)
- 路径2B:关键技能(从静态图片添加动效)
short/long-prompt-guide:
- 路径1:指导整个提示词结构
- 路径2B:指导动效/动画结构
imagine:
- 路径1:不使用
- 路径2A:构建Imagen提示词的核心指南
Workflow Decision Tree
工作流决策树
User: "I want to create a video"
↓
ARCH-V: "Text-to-Video or Image-to-Video?"
↓
Path 1: Text-to-Video
↓
"Short or Long prompt?"
↓
Load: great-prompt-anatomy + camera-movements + [short/long]-guide
↓
Validate 8 components + conflicts
↓
Output: Veo 3 prompt
Path 2: Image-to-Video
↓
STAGE A: "Create still image"
↓
Load: imagine + great-prompt-anatomy (visual components)
↓
Validate visual components
↓
Output: Imagen prompt
↓
User generates image
↓
STAGE B: "Add motion to image"
↓
"Short or Long motion?"
↓
Load: camera-movements + great-prompt-anatomy (motion components)
↓
Validate motion feasibility + conflicts
↓
Output: Veo 3 image-to-video prompt用户:"我想创建一个视频"
↓
ARCH-V:"选择文本转视频还是图片转视频?"
↓
路径1:文本转视频
↓
"选择短提示词还是长提示词?"
↓
加载:great-prompt-anatomy + camera-movements + [short/long]-guide
↓
验证8个组件 + 冲突检查
↓
输出:Veo 3提示词
路径2:图片转视频
↓
阶段A:"创建静态图片"
↓
加载:imagine + great-prompt-anatomy(视觉组件)
↓
验证视觉组件
↓
输出:Imagen提示词
↓
用户生成图片
↓
阶段B:"为图片添加动效"
↓
"选择短动效还是长动效?"
↓
加载:camera-movements + great-prompt-anatomy(动效组件)
↓
验证动效可行性 + 冲突检查
↓
输出:Veo 3图片转视频提示词Validation Rules
验证规则
Universal Validations (All Paths)
通用验证(所有路径)
Mandatory Component Presence:
- All required components for chosen path present
- No empty placeholders
- Specific details provided (not vague)
Camera Movement:
- ONE movement per beat/timestamp
- Uses standardized vocabulary from camera-movements skill
- Movement appropriate for shot type
Style Consistency:
- Style/aesthetic maintained throughout
- Color palette specified (3-5 color anchors)
- Lighting approach consistent
必填组件完整性:
- 所选路径要求的所有组件均已提供
- 无空占位符
- 提供具体细节(而非模糊描述)
镜头运动:
- 每个节拍/时间戳仅包含一种运动
- 使用camera-movements技能中的标准化术语
- 运动类型与镜头类型匹配
风格一致性:
- 全程保持风格/美学统一
- 指定调色板(3-5个颜色锚点)
- 光线处理方式一致
Path 1 Specific Validations
路径1专属验证
Time/Weather Continuity:
- Single time of day throughout (unless intentional transition)
- Weather consistent (no sudden rain → sun)
- Lighting quality matches time/weather
Audio Appropriateness:
- Dialogue formatted correctly ()
Character: "Text" - Ambient sounds match environment
- Music style fits mood
时间/天气连贯性:
- 全程使用单一时间段(除非是有意的过渡)
- 天气保持一致(无突然从下雨转晴天的情况)
- 光线质量与时间/天气匹配
音频适配性:
- 对话格式正确()
角色名: "台词" - 环境音与场景匹配
- 音乐风格符合氛围
Path 2 Specific Validations
路径2专属验证
Stage A (Imagen) Validations:
- No motion words (running, flying, moving) - image is static
- No audio descriptions - images silent
- Technical photography specs appropriate
- Token count under 480
Stage B (Veo 3 motion) Validations:
- Motion achievable from static starting point
- Camera movement respects image composition
- Action fits subject capabilities
- Audio matches visual aesthetic from Stage A
阶段A(Imagen)验证:
- 无运动类词汇(跑步、飞行、移动)- 图片为静态
- 无音频描述 - 图片无声音
- 专业摄影参数合适
- 令牌数量低于480
阶段B(Veo 3动效)验证:
- 动效应可从静态初始画面实现
- 镜头运动需尊重图片构图
- 动作需符合主体的能力
- 音频需匹配阶段A的视觉美学
Error Messages & Fixes
错误信息与修复方案
Common Issues
常见问题
Issue: Missing Components
⚠️ PROMPT-LOCKED
Missing mandatory components:
- Audio not specified
- Lighting/Mood undefined
Fix: Add audio description (dialogue/ambience/music)
Fix: Specify light sources and moodIssue: Camera Movement Conflict
⚠️ PROMPT-LOCKED
Conflict detected:
- "0-4s: Dolly in while panning left"
Fix: Choose ONE movement per beat
- Option A: "0-4s: Dolly in"
- Option B: "0-4s: Pan left"Issue: Time/Weather Conflict
⚠️ PROMPT-LOCKED
Conflict detected:
- Setting: "Golden hour sunset"
- Lighting: "Harsh midday sun"
Fix: Make lighting consistent with time
- "Golden hour: warm, low-angle sun, soft shadows"Issue: Path 2A - Motion in Static Image
⚠️ IMAGEN PROMPT LOCKED
Error: Motion detected in static image prompt
- "person running across field"
Fix: Describe static composition
- "person mid-stride in running pose"
Then add motion in Stage B (Veo 3)问题:缺失组件
⚠️ 提示词锁定
缺失必填组件:
- 未指定音频
- 未定义光线/氛围
修复方案:添加音频描述(对话/环境音/音乐)
修复方案:指定光源和氛围问题:镜头运动冲突
⚠️ 提示词锁定
检测到冲突:
- "0-4s: 推进镜头同时向左摇"
修复方案:每个节拍选择一种运动
- 选项A:"0-4s: 推进镜头"
- 选项B:"0-4s: 向左摇"问题:时间/天气冲突
⚠️ 提示词锁定
检测到冲突:
- 场景:"黄金时段日落"
- 光线:"正午强烈阳光"
修复方案:使光线与时间保持一致
- "黄金时段:温暖的低角度阳光,柔和阴影"问题:路径2A - 静态图片中包含运动
⚠️ Imagen提示词锁定
错误:静态图片提示词中包含运动描述
- "人物跑过田野"
修复方案:描述静态构图
- "人物处于跑步姿势的中间姿态"
然后在阶段B(Veo 3)中添加运动Example Workflows
工作流示例
Example 1: Path 1 Short Prompt
示例1:路径1短提示词
User: "I want a coffee shop morning scene"
ARCH-V: Path 1 (Text-to-Video) → Short Prompt
Guides through:
- Subject: Barista pouring latte art
- Setting: Morning café, 8:30 AM
- Action: Pours latte, heart shape forms
- Style: Cinematic B-roll, warm grade
- Camera: Overhead, slow tilt down
- Lighting: Natural window light, warm
- Audio: Espresso machine, milk steaming
- Constraints: (none specified)
Validates: ✅ All 8 components present
Output:
Format & style: Cinematic B-roll, warm color grade
Barista pours latte art, heart shape forming in foam
Morning café, 8:30 AM, soft window light
Overhead, slow tilt down
Natural window key, warm practical fills
Espresso machine hiss, milk steaming, gentle café chatter用户: "我想要一个咖啡馆早晨的场景"
ARCH-V: 路径1(文本转视频)→ 短提示词
引导内容:
- 主体:咖啡师制作拉花拿铁
- 场景:早晨的咖啡馆,8:30
- 动作:倒牛奶,拉出心形图案
- 风格:电影感B-roll,暖色调
- 镜头:俯拍,缓慢向下倾斜
- 光线:自然窗户光,温暖
- 音频:咖啡机声音,牛奶蒸汽声
- 约束条件:(无)
验证结果: ✅ 所有8个组件齐全
输出:
格式与风格:电影感B-roll,暖色调调色
咖啡师制作拉花拿铁,心形图案在奶泡中成型
早晨咖啡馆,8:30,柔和窗户光
俯拍,缓慢向下倾斜
自然窗户主光,温暖补光
咖啡机嘶嘶声,牛奶蒸汽声,轻柔咖啡馆 chatterExample 2: Path 1 Long Prompt
示例2:路径1长提示词
User: "Romantic rain scene with couple under umbrella"
ARCH-V: Path 1 (Text-to-Video) → Long Prompt (has dialogue)
Loads: long-prompt-guide Production Brief framework
Guides through all 11 blocks, validates:
- FG/MG/BG structure ✅
- Color anchors (5 colors) ✅
- Camera beats (3 beats, ONE movement each) ✅
- Dialogue format ✅
- Continuity rules ✅
Output: Complete Production Brief (see long-prompt-guide examples)
用户: "一对情侣在伞下的浪漫雨中场景"
ARCH-V: 路径1(文本转视频)→ 长提示词(包含对话)
加载: long-prompt-guide 制作简报框架
引导完成后验证:
- 前景/中景/背景结构 ✅
- 颜色锚点(5种颜色) ✅
- 镜头节拍(3个节拍,每个节拍一种运动) ✅
- 对话格式 ✅
- 连贯性规则 ✅
输出: 完整制作简报(详见long-prompt-guide示例)
Example 3: Path 2 Image-to-Video
示例3:路径2图片转视频
User: "I want precise control over Jakarta street scene, then animate it"
ARCH-V: Path 2 (Image-to-Video)
STAGE A - Imagen:
Subject: Young street vendor, simplified Science SARU style
Setting: Dense Jakarta street, golden hour
Style: Science SARU animation aesthetic
Camera: Wide establishing, low angle (STATIC)
Lighting: Golden hour warm, dramatic shadows
Constraints: No motion, frozen moment
[Validates: No motion words, visual only]
✅ IMAGEN PROMPT:
"Science SARU animation style aesthetic. Young Indonesian street
vendor character with simplified design, elastic limbs, large
expressive eyes, worn earth tone clothing. Dense layered Jakarta
street scene at golden hour, geometric shophouses, vibrant warm
palette transitioning to orange-pink sunset gradient, dramatic
long shadows, wet reflective surfaces. Wide establishing shot,
low angle, dramatic composition, flat color blocks with gradients,
painterly watercolor texture, melancholic atmosphere."User generates image in Imagen 3/4
STAGE B - Veo 3 Motion:
[Image uploaded to Veo 3]
Action: Vendor turns head slowly toward camera, slight smile
Camera: Slow dolly in from wide to medium close-up
Audio: Street ambience, distant traffic, vendor humming melody
[Validates: Motion achievable, camera respects composition]
✅ VEO 3 IMAGE-TO-VIDEO PROMPT:
"From this still image: Vendor turns head slowly toward camera,
slight smile forming. Slow dolly in from wide establishing to
medium close-up over 8 seconds. Jakarta street ambience with
distant traffic, motorbike sounds, vendor humming traditional
Indonesian melody. Maintain golden hour lighting, warm atmosphere,
and Science SARU aesthetic from image."用户: "我想精准控制雅加达街头场景,然后添加动画"
ARCH-V: 路径2(图片转视频)
阶段A - Imagen:
主体:年轻街头小贩,简化Science SARU风格
场景:密集的雅加达街头,黄金时段
风格:Science SARU动画美学
镜头:广角定场,低角度(静态)
光线:黄金时段暖光,戏剧性阴影
约束条件:无运动,定格瞬间
[验证结果:无运动词汇,仅包含视觉描述]
✅ Imagen提示词:
"Science SARU动画美学风格。年轻印尼街头小贩角色,简化设计,弹性四肢,大而有神的眼睛,磨损的大地色系服装。黄金时段的密集雅加达街头场景,几何形状的店铺,从暖色调到橙粉色日落渐变的活力调色板,戏剧性长阴影,潮湿反光的地面。广角定场镜头,低角度,戏剧性构图,带渐变的平色块,水彩画质感,忧郁氛围。"用户在Imagen 3/4中生成图片
阶段B - Veo 3动效:
[图片已上传至Veo 3]
动作:小贩缓慢转头看向镜头,露出微笑
镜头:从广角缓慢推进至中近景
音频:街头环境音,远处交通声,小贩哼歌
[验证结果:动效可实现,镜头尊重构图]
✅ Veo 3图片转视频提示词:
"基于此静态图片:小贩缓慢转头看向镜头,逐渐露出微笑。8秒内从广角定场镜头缓慢推进至中近景。雅加达街头环境音,远处交通声,摩托车声,小贩哼唱传统印尼旋律。保持黄金时段光线、温暖氛围和图片中的Science SARU美学。"Tips for Using ARCH-V
使用ARCH-V的技巧
Choose the right path:
- Path 1 faster for straightforward videos
- Path 2 better for complex visual compositions or when you need perfect still first
Be specific early:
- Detailed subject descriptions help ARCH-V guide you better
- Vague inputs → more back-and-forth questions
Trust the validation:
- PROMPT-LOCKED means real conflicts exist
- Suggested fixes based on proven patterns
- Resolve conflicts before generating
Use skill references:
- ARCH-V will load appropriate skills automatically
- You can reference them directly for inspiration
- Cross-skill integration is intentional
Iterate progressively:
- Start with minimum viable (4 blocks for long prompts)
- Add detail as needed
- ARCH-V guides what's optional vs mandatory
选择合适的路径:
- 路径1更适合制作简单直接的视频
- 路径2更适合复杂视觉构图或需要完美初始画面的场景
尽早提供具体细节:
- 详细的主体描述能帮助ARCH-V提供更精准的引导
- 模糊输入会导致更多来回确认的问题
信任验证机制:
- 提示词锁定意味着确实存在冲突
- 建议的修复方案基于经过验证的模式
- 生成前请解决所有冲突
使用技能参考:
- ARCH-V会自动加载合适的技能
- 你可以直接参考这些技能获取灵感
- 跨技能集成是经过精心设计的
逐步迭代优化:
- 从最小可行版本开始(长提示词先完成4个模块)
- 根据需要逐步添加细节
- ARCH-V会引导你区分可选和必填内容
Technical Notes
技术说明
Token Efficiency:
- ARCH-V itself: ~2,000 tokens
- Loads reference skills on-demand only
- Total system: ~8,600 tokens vs ~15,000 monolithic
- 43% token savings with full capability
Skill Integration:
- All 5 reference skills work together seamlessly
- Cross-references validated automatically
- No duplication between skills
Progressive Disclosure:
- ARCH-V SKILL.md loaded when user asks for video prompts
- Reference skills loaded based on path/choices
- Optimal token usage throughout workflow
令牌效率:
- ARCH-V自身:约2000令牌
- 仅按需加载参考技能
- 总系统令牌数:约8600令牌,相比单体系统的约15000令牌
- 在保持完整功能的前提下节省43%的令牌
技能集成:
- 所有5个参考技能无缝协作
- 自动验证交叉引用
- 技能间无重复内容
渐进式披露:
- 当用户请求视频提示词时加载ARCH-V SKILL.md
- 根据路径/选择加载参考技能
- 全程优化令牌使用
Quick Start
快速入门
For beginners: Let ARCH-V guide you with questions
For experienced users: Specify path and prompt type upfront:
- "Text-to-video short prompt for product shot"
- "Image-to-video, create Imagen prompt first for portrait"
For complex projects: Use Path 2 with long motion prompts for maximum control
ARCH-V adapts to your workflow preference!
新手用户: 让ARCH-V通过问题引导你完成
资深用户: 直接指定路径和提示词类型:
- "文本转视频短提示词,用于产品镜头"
- "图片转视频,先创建Imagen提示词用于人像"
复杂项目: 使用路径2和长动效提示词以获得最大控制
ARCH-V会适配你的工作流偏好!",