MiniMax H3 与 Grok 视频生成 | API 文档

# MiniMax H3 and Grok Video Generation

MiniMax H3 and Grok Imagine Video 1.5 asynchronous requests, limits, billing units, and job polling.

POST https://api.apihop.ai/v1/video-generations

## Availability

Availability depends on your API Hop API Key. Submit a request only when [GET https://api.apihop.ai/v1/models](https://apihop.ai/en/docs/api-reference/models) returns its exact model ID: `minimax-h3` or `grok-imagine-video-1.5`. Documentation does not enable access. Veo 3.1 Fast and Gemini Omni Flash are not available through this public guide.

## Request Parameters

Send JSON with `Authorization: Bearer YOUR_APIHOP_API_KEY` and `Content-Type: application/json`.

| model | seconds | resolution | images | audios |
| --- | --- | --- | --- | --- |
| `minimax-h3` | 6-30 | `768p` (default), `480p` | 0-8 | 0-3 |
| `grok-imagine-video-1.5` | 6-15 | `720p` | 0-7 | Unsupported |

`model` and a non-empty string `prompt` are required. Grok prompts may contain at most 1,200 characters. `seconds` accepts an integer or integer string and defaults to 6. `aspect_ratio` defaults to `16:9`; `9:16` is also accepted. `images` and `audios` are arrays of HTTP(S) URLs; Grok images also accept PNG, JPEG, or WebP base64 data URIs. H3 does not guarantee 480p output; use 768p. Omit `stream` or set it to `false`. Unsupported fields and conflicting aliases are rejected.

## Request Examples

```bash
curl --request POST \
  --url https://api.apihop.ai/v1/video-generations \
  --header 'Authorization: Bearer YOUR_APIHOP_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{"model":"minimax-h3","prompt":"A slow pan across a mountain lake","seconds":"6","resolution":"768p","aspect_ratio":"16:9"}'
```

```json
{
  "model": "grok-imagine-video-1.5",
  "prompt": "A slow pan across a mountain lake",
  "seconds": "6",
  "resolution": "720p",
  "aspect_ratio": "16:9"
}
```

## Job Status and Results

An accepted request returns HTTP `202`, `data.jobId` as a decimal string, and `data.status: "PENDING"`. Preserve the ID as a string. Poll [GET https://api.apihop.ai/v1/jobs/{id}](https://apihop.ai/en/docs/api-reference/jobs) with a key from the same account every 5-10 seconds. `PENDING`, `CONNECT`, and `PROCESSING` are non-terminal. On `COMPLETED`, read video URLs from `data.outputResult.result[].url`; each URL is a string. Stop polling on `FAILED` or `CANCELLED`. Save completed videos promptly. If submission times out or a job remains processing, check the existing job and account usage before retrying; another POST can create another charged task.

## Billing

MiniMax H3 is billed per second of generated video; Grok Imagine Video 1.5 is billed per request. See [Pricing](https://apihop.ai/en/docs/pricing) for published rates. Do not multiply Grok's per-request price by video duration.

可用性

模型可用性以你的 API Hop API Key 为准。仅当 GET /v1/models 返回对应的准确模型 ID（minimax-h3 或 grok-imagine-video-1.5）时，才提交该模型的请求。文档不代表已获得调用权限。Veo 3.1 Fast 与 Gemini Omni Flash 不在本公开指南的可用范围内。

请求参数

发送 JSON，并设置 Authorization: Bearer YOUR_APIHOP_API_KEY 与 Content-Type: application/json。

model	seconds	resolution	images	audios
`minimax-h3`	6-30	`768p`（默认）、`480p`	0-8	0-3
`grok-imagine-video-1.5`	6-15	`720p`	0-7	不支持

model 与非空字符串 prompt 必填。Grok 提示词最多 1,200 个字符。seconds 接受整数或整数字符串，默认为 6。aspect_ratio 默认为 16:9，也接受 9:16。images 与 audios 为 HTTP(S) URL 数组；Grok 图片也接受 PNG、JPEG 或 WebP 的 base64 data URI。H3 不保证输出 480p，建议使用 768p。省略 stream 或设为 false。不支持的字段与相互冲突的别名会被拒绝。

请求示例

bash

curl --request POST \
  --url https://api.apihop.ai/v1/video-generations \
  --header 'Authorization: Bearer YOUR_APIHOP_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{"model":"minimax-h3","prompt":"A slow pan across a mountain lake","seconds":"6","resolution":"768p","aspect_ratio":"16:9"}'

json

{
  "model": "grok-imagine-video-1.5",
  "prompt": "A slow pan across a mountain lake",
  "seconds": "6",
  "resolution": "720p",
  "aspect_ratio": "16:9"
}

任务状态与结果

请求接受后返回 HTTP 202、十进制字符串 data.jobId 和 data.status: "PENDING"。请将 ID 保持为字符串。使用同一账户的 Key，每 5-10 秒轮询 GET /v1/jobs/{id}。PENDING、CONNECT、PROCESSING 为未结束状态；COMPLETED 时读取 data.outputResult.result[].url，每个 URL 均为字符串。遇到 FAILED 或 CANCELLED 停止轮询。请及时保存生成视频。提交超时或任务持续处理时，重试前先检查原任务与账户用量；再次 POST 可能新建另一个收费任务。

计费

MiniMax H3 按生成视频的秒数计费；Grok Imagine Video 1.5 按次计费。已公开价格见 Pricing。不要将 Grok 的按次价格再乘以视频时长。