video-ai-direction
Expert guidance on building and prompting AI video sequences using the Kling API.
Install
mkdir -p .claude/skills/video-ai-direction && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/11828" && unzip -o skill.zip -d .claude/skills/video-ai-direction && rm skill.zipInstalls to .claude/skills/video-ai-direction
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
AI video generation direction for Kling API — prompt engineering, motion design, platform specs, text-to-video and image-to-video workflows for marketing contentKey capabilities
- →Construct text-to-video prompts using an 8-element structure
- →Generate video from static images with motion descriptions
- →Select Kling API models for different quality and speed needs
- →Poll for video generation completion using task IDs
- →Create three video variants for marketing content (hero, hook, CTA)
- →Apply platform-specific specifications for social media videos
How it works
This skill directs AI video generation using the Kling API by structuring prompts, selecting models, and applying platform specifications. It covers both text-to-video and image-to-video workflows.
Inputs & outputs
When to use video-ai-direction
- →Create marketing video prompts
- →Generate video from static assets
- →Optimize Kling API workflow
About this skill
Video AI Direction Skill
AI video generation using the Kling API (kling-v1-5 / kling-v2). This skill covers prompt construction, API workflow, platform specifications, and creative direction for marketing video content.
Kling API Architecture
Kling generates video asynchronously: you POST a generation request, receive a task ID, then poll until the video is ready.
Base URL: https://api.klingai.com/v1
Auth: Authorization: Bearer $KLING_API_KEY
Generation Modes
| Mode | Input | Best for |
|---|---|---|
text2video | Text prompt | Campaign concepts, abstract, motion-first |
image2video | Static image + motion prompt | Product animation, photo-to-video |
Model Selection (Deterministic)
| Model | Quality | Speed | Use |
|---|---|---|---|
kling-v1 | Good | Fastest | Testing, iterations |
kling-v1-5 | Better | Medium | Production ads (default) |
kling-v2 | Best | Slowest | Hero brand films |
Default for marketing: kling-v1-5 in pro mode.
Prompt Engineering
The 8-Element Prompt Structure
[OPENING_FRAME]: what the viewer sees in the first frame
[MOTION]: what moves, how it moves, camera behavior
[SUBJECT]: who/what is the protagonist
[STYLE]: cinematic / commercial / documentary / animated
[CAMERA]: static / dolly-in / pan / tilt / aerial / handheld
[LIGHTING]: natural / golden hour / studio / dramatic
[MOOD]: emotional register
[END_FRAME]: what the last frame holds
Negative prompt always includes:
blurry, distorted faces, watermark, text overlay, low quality, amateur, shaky, overexposed, artifacts
Motion Language Reference
Camera movements:
slow dolly in— intimate, focus drawgentle pan right— reveal, storytellingaerial overhead pull back— scale, contextstatic locked— stability, confidencehandheld slight movement— documentary authenticitysmooth tracking shot— following action
Subject motion:
subtle product rotation— product showcaseliquid pour in slow motion— food/beverage premiumhands interacting with [product]— demonstrationbackground bokeh shift— focus transitionsparticle/dust float— atmosphere
Platform Specifications (Deterministic)
tiktok_reels:
ratio: "9:16"
duration: "5" # 5s for ad, 10s for organic
hook_window: "2s"
key: mobile-first, native text overlays in post
linkedin:
ratio: "16:9"
duration: "5"
hook_window: "3s"
key: auto-plays muted, captions mandatory
instagram_feed:
ratio: "1:1"
duration: "5"
hook_window: "2s"
key: square for feed, 9:16 for reels
youtube_preroll:
ratio: "16:9"
duration: "5" # skip at 5s, make it unskippable
hook_window: "5s"
key: must deliver value before skip
twitter_x:
ratio: "16:9"
duration: "5"
hook_window: "2s"
key: auto-plays muted
API Call Patterns
Text-to-Video
curl -s -X POST https://api.klingai.com/v1/videos/text2video \
-H "Authorization: Bearer $KLING_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model_name": "kling-v1-5",
"prompt": "FULL_PROMPT",
"negative_prompt": "blurry, distorted faces, watermark, text overlay, low quality",
"cfg_scale": 0.5,
"mode": "pro",
"aspect_ratio": "16:9",
"duration": "5"
}'
Image-to-Video (animate a static image)
curl -s -X POST https://api.klingai.com/v1/videos/image2video \
-H "Authorization: Bearer $KLING_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model_name": "kling-v1-5",
"image": "IMAGE_URL_OR_BASE64",
"prompt": "MOTION_DESCRIPTION",
"negative_prompt": "blurry, distorted faces, watermark, low quality",
"cfg_scale": 0.5,
"mode": "pro",
"duration": "5"
}'
Poll for Completion
# Poll every 15 seconds, up to 5 minutes
TASK_ID="<returned from generation call>"
curl -s https://api.klingai.com/v1/videos/text2video/$TASK_ID \
-H "Authorization: Bearer $KLING_API_KEY"
# Check .data.task_status == "succeed" and extract .data.task_result.videos[0].url
Deliverable Structure
Every video brief produces 3 variants:
Variant 1 — Hero concept: Full prompt, primary creative vision Variant 2 — Hook variant: Different opening 2 seconds (stronger pattern interrupt) Variant 3 — CTA variant: Different ending (clearer or softer call-to-action frame)
Plus: script/caption copy for each variant (added in post via the video editor).
The 5-Second Rule
Every 5-second video must:
- Second 1–2: Interrupt — something unexpected, beautiful, or emotionally resonant
- Second 3–4: Communicate — one clear idea about the product or brand
- Second 5: Direct — a visual that implies the next step (product logo, CTA frame)
No exceptions for ad placements. For organic content the same structure applies, with more room for story in seconds 3–4.
Common Rationalizations
| Rationalization | Reality |
|---|---|
| "I'll use std mode for speed" | std mode produces visibly lower quality. Always use pro for client assets. |
| "The negative prompt isn't necessary" | Unguided generation consistently produces watermarks and text artifacts. Include it always. |
| "One variant is enough" | The first generation is rarely the strongest. Three variants surfaces the best option and gives the client choice. |
| "Platform ratio doesn't matter much" | Wrong ratio: 20–40% reach reduction on most platforms. Ratio is a technical requirement, not a preference. |
| "I'll caption later" | Caption brief belongs in the deliverable now. Captions double engagement on muted auto-play. |
When not to use it
- →When generating videos for purposes other than marketing content
- →When the Kling API is not available or authorized
- →When a single video variant is sufficient for the use case
Limitations
- →Requires an Authorization: Bearer $KLING_API_KEY
- →Kling generates video asynchronously
- →Negative prompt always includes specific terms
How it compares
This skill provides a structured, 8-element prompt engineering approach and specific workflow for the Kling API, unlike generic video generation tools.
Compared to similar skills
video-ai-direction side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| video-ai-direction (this skill) | 0 | 2mo | Review | Intermediate |
| jimeng-mcp-skill | 19 | 4mo | Caution | Intermediate |
| baoyu-article-illustrator | 18 | 2mo | No flags | Beginner |
| baoyu-cover-image | 10 | 2mo | No flags | Beginner |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
jimeng-mcp-skill
wwwzhouhui
使用jimeng-mcp-server进行AI图像和视频生成。当用户请求从文本生成图像、合成多张图片、从文本描述创建视频或为静态图像添加动画时使用此技能。支持四大核心能力:文生图、图像合成、文生视频、图生视频。需要jimeng-mcp-server在本地运行或通过SSE/HTTP访问。
baoyu-article-illustrator
JimLiu
Analyzes article structure, identifies positions requiring visual aids, generates illustrations with Type × Style two-dimension approach. Use when user asks to "illustrate article", "add images", "generate images for article", or "为文章配图".
baoyu-cover-image
JimLiu
Generates article cover images with 5 dimensions (type, palette, rendering, text, mood) combining 9 color palettes and 6 rendering styles. Supports cinematic (2.35:1), widescreen (16:9), and square (1:1) aspects. Use when user asks to "generate cover image", "create article cover", or "make cover".
ai-media
arcasilesgroup
Generates images, videos, and audio via AI models (fal-ai MCP): cheap iteration models, expensive production finals, cost-estimate before generation. Trigger for 'generate an image', 'create a thumbnail', 'make a voiceover', 'AI video', 'text to speech for'. Not for design composition; use /ai-visua
generate
tahatms12
You are a visual producer. Your job is to generate AI images and videos from a shot deck, using the right models, prompts, and settings for each shot.
video-creation
Sheldon-92
Professional video production judgment for AI coding agents — storytelling, motion design, audio, tools (HyperFrames/Remotion)