Nebula Veo Video Generation Quick Start

This document is intended for API users and describes how to invoke the Google Vertex Veo model to generate videos via the Nebula unified interface. It covers request formats, parameter descriptions, and common examples.

Supported models: veo-3.1-generate-preview, veo-3.1-fast-generate-preview, veo-3.0-generate-preview, veo-3.0-fast-generate-001.

1. Basic Information

Item	Content
Base URL	`https://llm.ai-nebula.com`
Endpoint Path	`POST /v1/video/generations`
Authentication Method	`Authorization: Bearer <API Key>`
Content-Type	`application/json`
Task Mode	Asynchronous: Submit Task → Poll Status → Get Result

2. Request Parameters

Field	Type	Required	Description
`model`	string	Yes	Model name: `veo-3.1-generate-preview`, `veo-3.1-fast-generate-preview`, `veo-3.0-generate-preview`, `veo-3.0-fast-generate-001`.
`prompt`	string	Yes	Text prompt describing the video content to be generated.
`durationSeconds`	int	No	Video duration (seconds), supports `4` / `6` / `8`, default `4`.
`aspectRatio`	string	No	Aspect ratio, only supports `16:9` or `9:16`, default `16:9`.
`resolution`	string	No	Resolution options, `720p` or `1080p`, default `1080p`.
`fps`	int	No	Frame rate, default 24, can be overridden as needed.
`sampleCount`	int	No	Number of videos generated per request, range `1-4`, default `1`.
`generateAudio`	bool	No	Whether to generate synchronized audio, default `false`; fast models will ignore this parameter and always output video with audio.
`personGeneration`	string	No	Person generation policy: `allow_all` (all ages) / `allow_adult` (adults only) / `dont_allow` (prohibited), default `allow_all`.
`addWatermark`	bool	No	Whether to add a watermark, default `false`.
`seed`	int	No	Random seed, fixing it allows for reproducible results.
`image`	string	No	Reference image for the first frame, supports HTTP(S) URL or Base64 Data URI.
`lastFrame`	string	No	Reference image for the last frame, supports HTTP(S) URL or Base64 Data URI.

Tip
If image is provided, Veo will use the first frame as the starting visual for transition; providing lastFrame simultaneously allows for first and last frame constraints.

3. Submission Examples

3.1 Curl

3.2 Python

4. Query Task

Endpoint: GET /v1/video/generations/{task_id}

Return Example (Success, sampleCount = 4):

{
 "task_id": "cHJvamVjdHMv...",
 "status": "succeeded",
 "format": "mp4",
 "url": [
  "https://nebula-ads.oss-cn-guangzhou.aliyuncs.com/2025/11/18/abc123/veo-demo-1.mp4",
  "https://nebula-ads.oss-cn-guangzhou.aliyuncs.com/2025/11/18/abc123/veo-demo-2.mp4",
  "https://nebula-ads.oss-cn-guangzhou.aliyuncs.com/2025/11/18/abc123/veo-demo-3.mp4",
  "https://nebula-ads.oss-cn-guangzhou.aliyuncs.com/2025/11/18/abc123/veo-demo-4.mp4"
 ],
 "metadata": {
  "durationSeconds": 12,
  "aspectRatio": "9:16",
  "resolution": "1080p",
  "fps": 24,
  "generateAudio": true,
  "response": {
   "videos": [
    {
     "mimeType": "video/mp4",
     "encoding": "mp4",
     "url": "https://nebula-ads.oss-cn-guangzhou.aliyuncs.com/2025/11/18/abc123/veo-demo-1.mp4"
    },
    {
     "mimeType": "video/mp4",
     "encoding": "mp4",
     "url": "https://nebula-ads.oss-cn-guangzhou.aliyuncs.com/2025/11/18/abc123/veo-demo-2.mp4"
    },
    {
     "mimeType": "video/mp4",
     "encoding": "mp4",
     "url": "https://nebula-ads.oss-cn-guangzhou.aliyuncs.com/2025/11/18/abc123/veo-demo-3.mp4"
    },
    {
     "mimeType": "video/mp4",
     "encoding": "mp4",
     "url": "https://nebula-ads.oss-cn-guangzhou.aliyuncs.com/2025/11/18/abc123/veo-demo-4.mp4"
    }
   ]
  }
 }
}

url field: A string when only 1 video is generated, an array of strings when multiple videos are generated.

sampleCount = 1: "url": "https://.../veo-demo-1.mp4"

sampleCount > 1: "url": ["https://.../veo-demo-1.mp4", "https://.../veo-demo-2.mp4", ...]

The metadata.response.videos array appends a url field, allowing direct access to the final address of each video segment.

Other fields (durationSeconds, aspectRatio, generateAudio, etc.) will echo the parameters used during invocation for easy tracking.

5. Common Input Combinations

Scenario	Required Fields	Optional Fields
Pure Text Generation	`model`, `prompt`	`durationSeconds`, `aspectRatio`, `resolution`, `generateAudio`
First Frame Reference	`model`, `prompt`, `image`	Other fields same as above
Start and End Frame Constraints	`model`, `prompt`, `image`, `lastFrame`	Same as above
High Resolution Generation	`model`, `prompt`, `resolution`, `aspectRatio`	`fps`, `generateAudio`

6. Response Status Description

Status	Description
`submitted` / `queued`	Task is queued, waiting to start.
`in_progress`	Task is processing, can be polled repeatedly.
`succeeded`	Generation successful, `url` is the final video address (string or string array).
`failed`	Task failed, `error` field provides error code and information.

7. Tips

It is recommended to poll the task status every 3~5 seconds to avoid triggering rate limits.

If url is a string array, please download individually according to your needs; a single string can be used directly for <video> playback or saved to disk.

If audio generation is needed, please set generateAudio to true, and the system will automatically select the billing tier that includes audio.

The default frame rate is 24; the fps parameter can be passed to override this if necessary.

If failed is returned and the message suggests a policy issue, try modifying the prompt or reducing the duration/resolution before retrying.

Fast models (veo-3.0-fast-generate-001, veo-3.1-fast-generate-preview) include audio by default and ignore the generateAudio parameter; they can be called directly as scenarios with audio.

Google Veo Video Model Interface Document

Nebula Veo Video Generation Quick Start#

1. Basic Information#

2. Request Parameters#

3. Submission Examples#

3.1 Curl#

3.2 Python#

4. Query Task#

5. Common Input Combinations#

6. Response Status Description#

7. Tips#