Video generation

Overview

AGTCloud provides multiple video-generation models and protocols. A video request usually creates a task first, then uses the task identifier to retrieve status and output.

Supported tasks

Task	Description	Main input
Text to video	Generate a video from a text description	Prompt
Image to video	Add motion to a still image	Prompt, image
Reference-based generation	Generate content from images or video supported by the model	Prompt, reference media
Model-specific extensions	Use start/end frames, camera controls, or other model features	Depends on the model and protocol

API access

Base URL

https://api.agtcloud.cn/v1

Text to video

curl 'https://api.agtcloud.cn/v1/video/generations' \
  -H 'Authorization: Bearer your-api-key' \
  -H 'Content-Type: application/json' \
  --data '{
    "model": "MODEL_ID",
    "prompt": "A train crosses a valley after the rain as the camera slowly moves forward"
  }'

Image to video

curl 'https://api.agtcloud.cn/v1/video/generations' \
  -H 'Authorization: Bearer your-api-key' \
  -H 'Content-Type: application/json' \
  --data '{
    "model": "MODEL_ID",
    "prompt": "Preserve the subject and have the person walk slowly toward the camera",
    "image": "https://example.com/input.jpg"
  }'

After creating a task, save its Task ID and retrieve it through the same protocol.

Retrieve a task

curl 'https://api.agtcloud.cn/v1/video/generations/TASK_ID' \
  -H 'Authorization: Bearer your-api-key'

See Create a video task and Retrieve a video task for the create and retrieve fields.

Choose a protocol

Protocol	Create endpoint	Follow-up
Platform Video Generations	`POST /v1/video/generations`	Retrieve output with the Task ID
OpenAI Videos	`POST /v1/videos`	Retrieve and download with the Video ID
Vendor-specific protocol	See the corresponding Reference	Use the task identifier and status defined by that protocol

Use the create, retrieve, and download endpoints from the same protocol for a given task.