Knowledge Base

Kie.ai is a unified AI API platform providing access to cutting-edge image, video, audio, and chat models at prices typically 30–50% lower than official APIs (up to 80% off for some models). This guide walks you through getting started, authenticating, creating tasks, and making the most of the platform.

📚 Table of Contents

1. Getting Started
   
2. Authentication
   
3. Available Models
   
4. How Tasks Work (Async Model)
   
5. Creating a Task
   
6. Checking Task Status
   
7. Image Models
   
8. Video Models
   
9. Audio Models
   
10. Chat Models
    
11. Rate Limits
    
12. Data Retention
    
13. Pricing & Credits
    
14. Tips & Best Practices
    
15. Support
    

🚀 Getting Started

1. Sign up at https://kie.ai
   
2. Get your API key at https://kie.ai/api-key
   
3. Browse available models at https://kie.ai/market
   
4. Test in the Playground — each model page has a built-in playground so you can experiment before calling the API
   
5. Check pricing at https://kie.ai/pricing
   
6. View your logs at https://kie.ai/logs
   

🔐 Authentication

Every API request must include these headers:

Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

⚠️ Security Warning: Never expose your API key in frontend code, browser apps, mobile apps, or public repositories (e.g. GitHub). Treat it like a password.

Additional protections available:

• Rate limits per key (hourly, daily, and total usage caps)
  
• IP whitelist — restrict API access to approved server IPs only
  

If you receive this error, check your headers:

{"code": 401, "msg": "You do not have access permissions"}

🗂️ Available Models

⚙️ How Tasks Work (Async Model)

All generation tasks on Kie.ai are asynchronous.

POST request  →  Task Created  →  Task ID returned
                                        ↓
                               Poll for status  OR  Receive webhook callback
                                        ↓
                               Task Complete → Get result URLs

✅ An HTTP 200 OK response only means your task was created — NOT that it's done.

You must either:

• Poll the task status using the taskId, OR
  
• Provide a callback URL (webhook) so Kie.ai POSTs results to you when done
  

📤 Creating a Task

Market API (Universal)

POST https://api.kie.ai/api/v1/jobs/createTask
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

Example body (Kling Image to Video):

{
  "model": "kling-v2.5-turbo/image-to-video-pro",
  "prompt": "A red blob spinning through a bright blue sky with fluffy clouds",
  "imageUrl": "https://example.com/your-image.jpg",
  "duration": 5,
  "callBackUrl": "https://your-server.com/webhook"
}

Response:

{
  "code": 200,
  "msg": "success",
  "data": {
    "taskId": "your-task-id-here"
  }
}

🔍 Checking Task Status

GET https://api.kie.ai/api/v1/jobs/recordInfo?taskId=YOUR_TASK_ID
Authorization: Bearer YOUR_API_KEY

Possible statuses: queuing → generating → success / fail

You can also view all task history at: https://kie.ai/logs

🖼️ Image Models

Typical cost: 10–50 credits per generation

🎬 Video Models

Typical cost: 100–500 credits per generation

Common video parameters:

• prompt — describe the scene/action (up to 1800 characters)
  
• imageUrl — source image for image-to-video
  
• duration — 5 or 10 seconds (some models support 15s)
  
• quality — 720p or 1080p
  
• aspectRatio — 16:9, 9:16, 1:1, 4:3, 3:4
  

🎵 Audio Models

ElevenLabs (via Kie.ai)

Suno (Music)

💬 Chat Models

🚦 Rate Limits

If you exceed limits, you'll receive HTTP 429. Rejected requests do NOT enter the queue — they must be retried. Contact support if you consistently hit rate limits and need them raised.

🗄️ Data Retention

⚠️ Download and store your generated files before they expire!

💰 Pricing & Credits

• Prices are 30–50% cheaper than official provider APIs
  
• Some models offer up to 80% discount
  
• Check your remaining credits:
  

GET https://api.kie.ai/api/v1/chat/credit
Authorization: Bearer YOUR_API_KEY

Full pricing: https://kie.ai/pricing

💡 Tips & Best Practices

1. Write detailed prompts — More specific descriptions produce better results. Include scene details, camera movements, lighting, style, and effects.
   
2. Use the Playground first — Test model behaviour before integrating into your code.
   
3. Don't poll too fast — Wait a few seconds between status checks to avoid unnecessary requests.
   
4. Use webhooks in production — Callbacks (webhook URLs) are more efficient than polling.
   
5. Download files promptly — Generated files are deleted after 14 days.
   
6. Keep your API key secure — Use environment variables, never hardcode it.
   
7. Use IP whitelisting — Lock down your API key to known server IPs for extra security.
   
8. Check logs for debugging — Visit https://kie.ai/logs for full task history including credit usage.
   

🛟 Support

Support hours: UTC 21:00 – UTC 17:00 (next day)

🔗 Useful Links

Last updated: May 2026