Image-to-video

​

Wan video generation

​

Quick flow

1. Submit job → get task_id
2. Poll status → wait until completed for a public URL

​

Basics

• Endpoint: POST /v1/video/generations
• Auth: Bearer Token (Authorization: Bearer sk-xxxxxx)
• Content-Type: application/json

​

Request body

​

Response

​

Request example

{
"model": "wan2.6-i2v-flash",
"prompt": "A kitten running on grass, cinematic look, natural light and shadow",
"input_reference": "https://images.unsplash.com/photo-1514888286974-6c03e2ca1dba?w=1024",
"duration": 5
}

​

Response example

{
"id": "task_xGOhEWZIx3qiVkqL7rCE0np45eaTM2HA",
"task_id": "task_xGOhEWZIx3qiVkqL7rCE0np45eaTM2HA",
"object": "video",
"model": "wan2.6-i2v-flash",
"status": "queued",
"progress": 0,
"created_at": 1775026176
}

​

Notes

1. Image quality caps how good the video can look—prefer sharp, clear framing.
2. Async: Save task_id and poll the status API or use callbacks.
3. Prompts: Focus on motion and atmosphere, not restating static objects in the frame.
4. Environments: Example points at http://192.168.5.29:4000 for testing—use production host when live.