ai-image-video-generation（AI 图像视频生成）

VibeVideo Generation Skill

Prerequisites

• Environment variable VIBEVIDEO_API_KEY must be set with a valid API key
• Get your API key from: Dashboard → Settings → API Keys

API Endpoint

This skill always uses the official VibeVideo API endpoint: https://vibevideo.app

Generate Image

Create an image generation task:

curl -s -X POST https://vibevideo.app/api/ai/generate \
  -H "Authorization: Bearer $VIBEVIDEO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "mediaType": "image",
    "scene": "text-to-image",
    "model": "nano-banana-2",
    "prompt": "A cat sitting on a rainbow",
    "options": {
      "aspect_ratio": "1:1",
      "quality": "2K"
    }
  }'

For image-to-image, set "scene": "image-to-image" and add "image_url": "..." in options.

Response:

{ "code": 0, "data": { "id": "task_id", "status": "pending", "taskId": "...", "costCredits": 5 } }

Generate Video

Create a video generation task:

curl -s -X POST https://vibevideo.app/api/ai/generate \
  -H "Authorization: Bearer $VIBEVIDEO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "mediaType": "video",
    "scene": "text-to-video",
    "model": "seedance-2-0",
    "prompt": "A dog playing in a park",
    "options": {
      "resolution": "720p",
      "duration": "5s",
      "aspect_ratio": "16:9"
    }
  }'

For image-to-video, set "scene": "image-to-video" and add "image_url": "..." in options. For frames-to-video, add "start_image_url": "..." and "end_image_url": "..." in options.

Query Task Status

Tasks are asynchronous. Poll until status is success, failed, or canceled:

curl -s -X POST https://vibevideo.app/api/ai/query \
  -H "Authorization: Bearer $VIBEVIDEO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "taskId": "YOUR_TASK_ID" }'

Response includes status, taskInfo, taskResult, and taskUrls (JSON string of media URLs).

Calculate Cost

Check credit cost before generating:

curl -s -X POST https://vibevideo.app/api/ai/cost \
  -H "Authorization: Bearer $VIBEVIDEO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2-0",
    "mediaType": "video",
    "scene": "text-to-video",
    "options": { "resolution": "720p", "duration": "5s" }
  }'

Cancel Task

curl -s -X DELETE https://vibevideo.app/api/ai/tasks/YOUR_TASK_ID \
  -H "Authorization: Bearer $VIBEVIDEO_API_KEY"

Workflow

1. If the user doesn't specify a model, pick the default for the scene (see tables below)
2. Call the generate endpoint
3. Poll the query endpoint every 5 seconds until status is terminal (success/failed/canceled)
4. Parse taskUrls from the response and report the media URL(s) to the user
5. If code is not 0, handle the error (see Error Handling below)

Image Models

Default for text-to-image: nano-banana-2

Video Models

Default for text-to-video: seedance-2-0

Error Handling

• code: -1 with "no auth": User's API key is missing or invalid. Remind them to set VIBEVIDEO_API_KEY.
• code: -1002 with "insufficient credits": User needs to purchase credits at VibeVideo dashboard.
• code: -1 with "invalid": Wrong model ID, scene, or mediaType. Check against the model tables above.
• Task stuck in "processing": Polling timeout. The task may still complete — suggest the user wait and query again later.

API Response Envelope

All endpoints return:

{ "code": 0, "message": "ok", "data": { ... } }

code: 0 means success. Non-zero code means error (check message).