Skip to content

Latest commit

 

History

History
102 lines (90 loc) · 3.12 KB

KIE_GrokImagine_T2V_Spec.md

File metadata and controls

102 lines (90 loc) · 3.12 KB

KIE Grok Imagine (T2V) API Spec

Status

Reference spec for the implemented Grok Imagine text-to-video node. See KIE_GrokImagine_T2V.md for the ComfyUI node surface.

Endpoint

  • Method: POST
  • Path: /api/v1/jobs/createTask
  • Base URL: https://api.kie.ai
  • Model: grok-imagine/text-to-video

Request Body

{
  "model": "grok-imagine/text-to-video",
  "callBackUrl": "https://your-domain.com/api/callback",
  "input": {
    "prompt": "A couple of doors open to the right one by one randomly and stay open, to show the inside, each is either an living room, or a kitchen, or a bedroom or an office, with little people living inside.",
    "aspect_ratio": "2:3",
    "mode": "normal",
    "duration": "6",
    "resolution": "480p"
  }
}

Root Parameters

  • model (STRING, required): Must be grok-imagine/text-to-video.
  • callBackUrl (STRING, optional): Receives task completion notifications when provided.
  • input (OBJECT, required): Generation parameters for the model.

Input Parameters

  • prompt (STRING, required): Motion prompt for the generated video. Max length: 5000.
  • aspect_ratio (STRING, optional): One of 2:3, 3:2, 1:1, 9:16, 16:9.
  • mode (STRING, optional): One of fun, normal, spicy.
  • duration (STRING, optional): One of 6, 10, 15 seconds.
  • resolution (STRING, optional): One of 480p, 720p.

Success Response

{
  "code": 200,
  "message": "success",
  "data": {
    "taskId": "task_12345678"
  }
}

Callback Behavior

If callBackUrl is present, KIE posts task completion payloads to that URL for both success and failure states.

Success Callback Fields

  • code
  • data.completeTime
  • data.costTime
  • data.createTime
  • data.model
  • data.param
  • data.resultJson
  • data.state
  • data.taskId
  • data.failCode
  • data.failMsg
  • msg

Failure Callback Fields

  • code
  • data.completeTime
  • data.costTime
  • data.createTime
  • data.failCode
  • data.failMsg
  • data.model
  • data.param
  • data.state
  • data.taskId
  • data.resultJson
  • msg

ComfyUI Mapping Notes

Recommended node shape for this repo:

  • prompt: STRING multiline, required.
  • aspect_ratio: COMBO, optional.
  • mode: COMBO, optional.
  • duration: COMBO, optional.
  • resolution: COMBO, optional.
  • log: BOOLEAN, optional.
  • Output: VIDEO.

This fits the same helper pattern already used by Kling and Seedance text-to-video nodes:

  • validate prompt and enum inputs before submission
  • create task through kie_api/jobs.py
  • poll through kie_api/jobs.py
  • extract resultUrls through kie_api/results.py
  • download video and convert through kie_api/video.py
  • log credits through kie_api/credits.py

Implementation Notes

  • callBackUrl should stay transport-level and does not need to be exposed in the first ComfyUI node pass.
  • Internal polling defaults can follow the current async video node pattern in this repo.
  • The provided callback success example uses a .jpg URL in resultUrls, which conflicts with the endpoint type. Treat that as a documentation inconsistency unless live API behavior proves otherwise.