Note: If this is your first time using it, make sure videoclaw is installed:
uvx videoclaw --help
Trigger Scenarios
Triggered when the user says "make an xx video" or "help me generate a video."
Process Overview
1. 故事大纲(用户描述 → AI 整理)
2. 素材准备(AI 推断需求 → 用户确认 → 素材处理)
3. 文本分镜(Seedance 2.0 格式)
4. i2v 生成视频
Detailed Process
Step 1: Story Outline
Use AskUserQuestion to ask the user:
"Please describe the video idea you want to make (e.g., an astronaut discovers alien ruins on Mars)"
- User inputs idea
Use the following prompt to generate the story outline:
你是一个专业的视频策划专家。请根据用户的想法生成一个简化的故事大纲。
用户想法:<用户输入>
请生成以下格式的故事大纲:
【视频主题】一句话概括视频核心内容
【核心剧情】2-3句话描述视频的主要情节
【角色设定】
- 角色1:描述(包括外貌、性格)
- 角色2:描述
(如果没有明确角色,可省略)
【时长】X秒(默认15秒)
【风格偏好】用户指定的风格,如果没有则留空
After generation, save to <project>/outline.md
Use AskUserQuestion to ask:
"The story outline is as follows. Are you satisfied?"
- Satisfied → Continue to Step 2
- Adjust → Ask what specifically needs to be changed
Step 2: Asset Preparation
2.1 Infer Asset Requirements
Infer which assets are needed based on the story outline:
基于以下故事大纲,分析需要哪些素材:
故事大纲:
<故事大纲内容>
请列出:
【角色素材】
- 需要几个角色?
- 每个角色需要什么类型的图片?
【场景素材】
- 需要什么场景?
- 是否需要背景图?
【参考素材】(可选)
- 是否需要视频参考?(运镜/动作)
- 是否需要音频参考?(配乐/对白)
请用清晰的列表格式回复。
2.2 User Confirms Assets
Use AskUserQuestion to ask:
"Based on the story outline, it is recommended to prepare the following assets:
- Characters: xxx
- Scenes: xxx
- Reference video: optional
- Reference audio: optional
Do you have these assets? What needs to be added?"
- Confirm existing assets
- Add new assets
- Skip certain assets
2.3 Asset Processing
For each required asset:
Character image (nine-grid): A qualified character asset needs to include the following views, placed in a single nine-grid image:
- Front, side, back, 3/4 side
- Facial close-up
- Close-ups of multiple expressions
- If there is a weapon: multiple views of the weapon
- If there are accessories (hats, etc.): image without the hat + separate image of the accessory
Character Image Generation:
- User provides → Skip generation
- No reference image → Generate in T2I mode (use video-t2i skill)
- Has reference image → Generate in I2I mode (use video-i2i skill)
Example T2I prompt for generating a character nine-grid:
一个九宫格图片,展示一个黄色圆滚滚小怪兽角色设计。
左上格:正面全身照
上中格:3/4侧面全身照
右上格:侧面全身照
左中格:正面表情-微笑
中格:正面表情-严肃
右中格:正面表情-惊讶
左下格:背面全身照
下中格:面部特写
右下格:正面表情-愤怒
风格:卡通形象,无鼻设计,头顶一个小橘子
Example I2I prompt for generating a character nine-grid:
保持角色特征不变,生成九宫格格式的角色展示图。
包含:正面、侧面、背面、3/4侧面、面部特写、多种表情
Scene image:
- User provides → Skip
- Needs generation → Generate in T2I mode (use video-t2i skill)
After generation, save to <project>/assets/
Use AskUserQuestion to confirm:
"The assets are ready. Are you satisfied?"
Step 3: Text Storyboard
Use the story-type storyboard template to generate the text storyboard. See details in video-text-storyboard/references/stories.md:
基于以下故事大纲和素材,使用故事类分镜模板生成文本分镜。
故事大纲:
<故事大纲内容>
素材:
- 角色图:<路径>
- 场景图:<路径>
- 参考视频:<路径>(如有)
- 参考音频:<路径>(如有)
请使用以下格式:
【素材介绍】:
【@图1】<角色/物体描述>
【@图2】<角色/物体描述>
【场景描写】:
<整体背景描述>
【背景音】:
<背景音描述>
【分镜】:
【镜头标题】(X-Y秒)
画面:<包含动作、台词、情绪的完整描述>
音效:<此镜头的音效>
【注意事项】:
- 角色设定:<关键角色特征>
- 氛围:<整体氛围要求>
- 动作节奏:<节奏要求>
- 细节还原:<关键细节>
After generation, save to <project>/videos/<video>/storyboard/text_storyboard.md
Use AskUserQuestion to confirm:
"The text storyboard is as follows. Are you satisfied?"
- Satisfied → Continue to Step 4
- Adjust → Ask what specifically needs to be changed
Step 4: i2v Video Generation
Execute the i2v command:
videoclaw i2v -p <project> \
-i <角色图> \
-i <场景图> \
--video-ref <参考视频> \
--audio-ref <参考音频> \
-t "<文本分镜内容>"
After generation, use AskUserQuestion to ask:
"The video has been generated. Are you satisfied?"
- Satisfied → End
- Regenerate → Retry
- Adjust → Modify and retry
Storyboard Format Reference
See details in video-text-storyboard/references/stories.md
【素材介绍】:
【@图1】黄色圆滚滚小怪兽,无鼻设计,头顶小橘子
【@图2】冰封圣剑,剑身蓝光流转
【@图3】冰之王冠,由冰晶碎片拼接而成
【@图4】冰封王座,台阶和座椅有冰晶纹路
【@图5】极寒冰封王座大殿场景,浮空冰峰刺破阴云
【场景描写】:
极寒冰封之地,浮空冰峰刺破阴云,寒风呼啸,冰封大殿寂静威严
【背景音】:
史诗感 夹杂冰面细微碎裂声
【分镜】:
【特写圣剑】(0-4秒)
画面:特写【@图2】冰封圣剑,剑身蓝光流转,插在寒冰基座上
音效:冰裂声、低沉嗡鸣
【拔剑】(4-8秒)
画面:镜头后拉。【@图1】黄色小怪兽肉垫爪子紧紧握住【@图2】剑柄,用力向上拔剑
音效:拔剑声、冰裂声
【冰甲覆身】(8-12秒)
画面:全身镜头。【@图1】小怪兽拔出圣剑瞬间,无数冰晶碎片从四面八方飞聚而来,瞬间拼接成完整冰甲覆盖全身
音效:剑鸣、冰晶碰撞声
【戴上王冠】(12-16秒)
画面:【@图1】小怪兽抬手,冰晶拼成【@图3】王冠;将王冠戴在头上
音效:冰晶拼接声
【登基定格】(16-20秒)
画面:全景。【@图1】小怪兽手持【@图2】圣剑站在【@图4】王座上,双眼亮起冰蓝光芒,远处【@图5】冰峰堡垒矗立
音效:余音回荡
【注意事项】:
- 角色设定:黄色小怪兽严格保持无鼻设计,头顶小橘子
- 氛围:全程冷色调,突出寒冰、冰晶质感
- 动作节奏:每个镜头一个简单动作,拔剑 → 冰甲覆身 → 戴王冠 → 登基
Common Mistakes
❌ Chaotic material references: not specifying what is referenced
✅ Use the 【@Image 1】【@Image 2】 format to explicitly reference materials
❌ Content overload: cramming too many scenes into 4-5 seconds
✅ Streamline content: arrange only one simple action per shot
❌ Ignoring sound effects: sound effect design can greatly improve quality
✅ Add sound effects: background music + sound effects
❌ Missing notes: the ending does not emphasize key elements
✅ Add notes: character settings, atmosphere, action rhythm, detail restoration