Video Generation | ModKie Docs

Generate short AI videos through the site API — no web UI required. This covers direct AI video generation (text-to-video, image-to-video, etc.), not workflow-based video pipelines.

Configuration details: CLI & API Guide.

Prerequisites

Create a key at Settings → API Keys (sk-...)
Configure the CLI:

pnpm ads login --host https://modkie.com --api-key sk-xxxxxxxx

Video generation consumes more credits than images. Ensure your account has sufficient balance.

Commands

Generate and wait

pnpm ads video generate \
  --prompt "cinematic product shot, soft lighting, slow camera push" \
  --scene text-to-video \
  --out ./output/video.mp4

Optional: --timeout 600000 (milliseconds, default 600000).

Image-to-video

pnpm ads video generate \
  --prompt "gentle camera motion, natural lighting" \
  --scene image-to-video \
  --image-input "https://example.com/reference.jpg" \
  --out ./output/motion.mp4 \
  --json

Query a task

pnpm ads video query --task-id <uuid> --json

List capabilities / schema

pnpm ads actions list
pnpm ads actions schema ai.video.generate_and_wait

Scenes

scene	Description
`text-to-video`	Generate from text prompt (default)
`image-to-video`	Animate a reference image
`video-to-video`	Transform an existing video
`ad-video`	Marketing ad video scene

Default model

When --provider and --model are omitted:

text-to-video: kie / bytedance/seedance-2-mini
image-to-video: kie / wan/2-7-image-to-video

Override with --provider and --model for other catalog models.

CLI flags vs API fields

Flag	API field
`--prompt`	`prompt`
`--scene`	`scene`
`--provider`	`provider`
`--model`	`model`
`--image-input`	`imageInput` (comma-separated)
`--video-input`	`videoInput` (comma-separated)
`--timeout`	`timeoutMs`
`--out`	downloads first `videoUrl` locally

API-only fields: options, workspaceId.

`--json` output example

{
  "taskId": "...",
  "status": "success",
  "videoUrl": "https://...",
  "localPath": "./output/video.mp4",
  "url": "https://...",
  "provider": "kie",
  "model": "bytedance/seedance-2-mini"
}

HTTP examples

Generate (blocking):

curl -H "Authorization: Bearer sk-xxx" \
  -H "Content-Type: application/json" \
  -X POST https://modkie.com/api/actions/ai.video.generate_and_wait \
  -d '{"prompt":"ocean waves at sunset","scene":"text-to-video","timeoutMs":600000}'

Async create + poll:

curl -H "Authorization: Bearer sk-xxx" \
  -H "Content-Type: application/json" \
  -X POST https://modkie.com/api/actions/ai.video.generate \
  -d '{"prompt":"ocean waves at sunset","scene":"text-to-video"}'

curl -H "Authorization: Bearer sk-xxx" \
  -H "Content-Type: application/json" \
  -X POST https://modkie.com/api/actions/ai.video.query \
  -d '{"taskId":"<uuid>"}'

API endpoints

Method	URL
`GET`	`https://modkie.com/api/actions`
`GET`	`https://modkie.com/api/actions/ai.video.generate_and_wait/schema`
`POST`	`https://modkie.com/api/actions/ai.video.generate_and_wait`
`POST`	`https://modkie.com/api/actions/ai.video.query`

Auth: Authorization: Bearer sk-xxx or X-Api-Key: sk-xxx

Notes

Video tasks often take 2–10 minutes. Use async ai.video.generate + ai.video.query for long runs.
Workflow video nodes and stock-clip compose are separate capabilities (not covered here).

On this page