跳转到内容

H3 Max Camera Controls

集成

模型 id minimax/h3-max/camera-controls
# Run the model; wait up to 20 s for the finished result in the same call
curl -X POST "https://api.doitong.com/v1/run/minimax/h3-max/camera-controls?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt_expansion_mode": "balanced",
    "image": "https://content.doitong.com/images/showcase/v3/gen_fox_cartoon.jpg"
  }'

# Still "queued" or "running"? Poll with the id from the response:
curl "https://api.doitong.com/v1/requests/REQUEST_ID?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY"

此请求: $0.163 163 API 积分 在此页面运行

获取 API 密钥

在 Doitong 上制作

Doitong 社区使用 Hailuo 03 New Turbo 制作的公开作品。一键复用提示词。

显示的公开作品:16
Close-up shot of a striking model with piercing eyes and flawless skin, a white snake elegantly c... 视频

Close-up shot of a striking model with piercing eyes and flawless skin, a white snake elegantly coiled around her head and neck, its scales glistening in soft, diffused light. The camera gently zooms in, emphasizing the intricate texture of the snake’s scales and the model’s striking facial...

Animate the stickman drawings into a smooth dance animation. The simple black stick figure perfor... 视频

Animate the stickman drawings into a smooth dance animation. The simple black stick figure performs fun rhythmic dance moves — waving its arms, bending knees, jumping lightly, spinning, and swaying side to side in time with upbeat music. Keep the white background clean and minimal. The motion...

A hyper-realistic, cinematic transformation sequence of the man in the uploaded image, sitting in... 视频

A hyper-realistic, cinematic transformation sequence of the man in the uploaded image, sitting in meditation inside a traditional Japanese room. The atmosphere begins to tremble slightly as soft blue light flows around him. From the ground, thousands of magnetic nano-particles rise, shimmering...

A cinematic transformation video of the man. The scene begins with him standing naturally, then l... 视频

A cinematic transformation video of the man. The scene begins with him standing naturally, then looking down at his hands as glowing holographic projections appear around his body. Mechanical parts of the Iron Man armor start assembling piece by piece with smooth animation: chest plate forming,...

An ultra-epic anime cinematic of Luffy’s greatest enemy unleashing a terrifying blind rage, rampa... 视频

An ultra-epic anime cinematic of Luffy’s greatest enemy unleashing a terrifying blind rage, rampaging uncontrollably against nature itself. His gigantic body is surrounded by a storm of black and crimson energy, crackling with lightning that scorches the ground. His roar shakes the heavens,...

A fair-skinned red-haired woman with freckles tightly embraces a majestic white tiger with faint ... 视频

A fair-skinned red-haired woman with freckles tightly embraces a majestic white tiger with faint reddish stripes. They remain close, perfectly synchronized in stillness. The woman blinks slowly, slightly tilting her head with a serious and intense expression. The tiger breathes gently — its...

The snowboarder is riding across massive, fluffy white clouds as if they were snow-covered mounta... 视频

The snowboarder is riding across massive, fluffy white clouds as if they were snow-covered mountains, with dramatic lighting and deep shadows adding volume and realism. The background shows a stunning blue sky with scattered soft clouds and subtle light rays. crisp cinematic quality, vivid...

Мрачная комната для допросов, тесная и давящая. Одна резкая лампа под потолком бросает конус холо... 视频

Мрачная комната для допросов, тесная и давящая. Одна резкая лампа под потолком бросает конус холодного света на металлический стол, вокруг густые тени. Виновный подозреваемый сидит в наручниках по одну сторону стола: потеет, нервно бегает глазами, тяжело дышит, избегает взгляда. Напротив —...

构建请求

控件根据模型架构生成。更改控件时,代码和价格会同步更新。

0 / 50000

覆盖默认静止场景行为的视频生成提示词。留空或省略会保持参考场景固定,仅允许摄像机移动。

•必填

生成前的提示词改写强度:'quality' 会在最长约 30s 内生成更丰富的提示词,而 'balanced' 大约在一秒内返回。选择 'disabled' 可跳过提示词扩展。

•必填 image_url
拖放文件,或

定义视频开头帧的图像 URL。

resolution 默认 480P

生成分辨率选择:1080P 需要对原生以 768P 生成的内容进行潜空间细化,而不是以所选分辨率原生生成。

默认 5

请求的时长(秒),实际视频长度最多可超出该请求约 0.7 秒。

高级 seed · enable_safety_checker · camera_trajectory

随机生成的种子;省略时会随机选择。

按顺序排列的摄像机关键帧,保留带符号的完整旋转,总方位角移动限制为 32 圈。在关键帧时间范围之外,摄像机会在首个时间戳之前保持第一姿态,并在视频结束前保持最终姿态。

POST /run/minimax/h3-max/camera-controls?wait=20 报价 → $0.163
# Run the model; wait up to 20 s for the finished result in the same call
curl -X POST "https://api.doitong.com/v1/run/minimax/h3-max/camera-controls?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt_expansion_mode": "balanced",
    "image": "https://content.doitong.com/images/showcase/v3/gen_fox_cartoon.jpg"
  }'

# Still "queued" or "running"? Poll with the id from the response:
curl "https://api.doitong.com/v1/requests/REQUEST_ID?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY"

运行此请求

你的密钥会保留在此浏览器中:仅发送到 Doitong API;只有勾选“记住”后才会保存到 localStorage。 获取 API 密钥

关于 H3 Max Camera Controls

将 `image_url`用作首帧,并沿 `camera_trajectory`移动摄影机;该参数是包含 2 至 12 个有序关键帧的列表。默认情况下场景保持刚性,仅摄影机发生变化。时长为 0.92 至 15 秒,分辨率为 480P、768P 或 1080P。可调节提示词扩展。

参数

发送规范名称或原生名称之一,不要同时发送两者。 规范名称

名称类型 默认值允许值 描述
prompt可选
类型string
默认值The same elements in the reference are rigid. Preserve every element exactly. The entire scene is frozen. Only the camera moves. no scene motion only camera motion
≤ 50,000 个字符
覆盖默认静止场景行为的视频生成提示词。留空或省略会保持参考场景固定,仅允许摄像机移动。
duration可选
类型number
默认值5
0.92 – 15
请求的时长(秒),实际视频长度最多可超出该请求约 0.7 秒。
resolution可选
类型enum
默认值480P
480P768P1080P
生成分辨率选择:1080P 需要对原生以 768P 生成的内容进行潜空间细化,而不是以所选分辨率原生生成。
seed可选
类型integer | null
随机生成的种子;省略时会随机选择。
enable_safety_checker可选
类型boolean
默认值true
安全检查器激活状态;true 启用检查。
prompt_expansion_mode必填
类型string
默认值balanced
生成前的提示词改写强度:'quality' 会在最长约 30s 内生成更丰富的提示词,而 'balanced' 大约在一秒内返回。选择 'disabled' 可跳过提示词扩展。
image_url必填 image
类型url · image
定义视频开头帧的图像 URL。
camera_trajectory可选
类型object[]
≤ 12 个项目≥ 2 个项目
按顺序排列的摄像机关键帧,保留带符号的完整旋转,总方位角移动限制为 32 圈。在关键帧时间范围之外,摄像机会在首个时间戳之前保持第一姿态,并在视频结束前保持最终姿态。
camera_trajectory[].time必填
类型number
0 – 1
视频内的归一化时间,使用区间 [0, 1] 分别表示开头和结尾。
camera_trajectory[].azimuth必填
类型number
摄像机相对于主体的水平角度,以度为单位。
camera_trajectory[].elevation必填
类型number
-90 – 90
摄像机相对于主体的垂直角度,以度为单位。
camera_trajectory[].distance必填
类型number
> 0
摄像机与主体之间的距离,以归一化场景单位表示。
必填 规范名称,也可接受

文件参数接受公开的 https:// URL、最大 10 MB 的 data: URI,或上传文件的 file_url。 文件

输出

GET /requests/req_8k2m4q7x1z9c3v5b6n0p2r4t
{
  "id": "req_8k2m4q7x1z9c3v5b6n0p2r4t",
  "object": "request",
  "model": "minimax/h3-max/camera-controls",
  "status": "succeeded",
  "created_at": "2026-10-20T12:00:00Z",
  "started_at": "2026-10-20T12:00:02Z",
  "completed_at": "2026-10-20T12:00:09Z",
  "queue_position": null,
  "cost": {
    "credits": 163,
    "usd": "0.163",
    "charged": 163,
    "refunded": 0
  },
  "input": {
    "image_url": "https://content.doitong.com/images/showcase/v3/gen_fox_cartoon.jpg",
    "prompt_expansion_mode": "balanced"
  },
  "output": {
    "video": {
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.mp4",
      "content_type": "video/mp4",
      "file_name": "0-output.mp4",
      "file_size": 1843200
    },
    "expanded_prompt": "string",
    "timings": {}
  },
  "outputs": [
    {
      "type": "video",
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.mp4",
      "content_type": "video/mp4",
      "file_size": 1843200,
      "width": 1280,
      "height": 720,
      "duration": 5,
      "stored": true
    }
  ],
  "error": null,
  "metrics": {
    "inference_seconds": 6.8
  },
  "urls": {
    "self": "https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t",
    "cancel": "https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t/cancel"
  }
}
你的文件,由你选择每个 outputs[].url 都是存储中的副本:可永久保存(按 GB-月付费),也可免费保存 7 天,由账户设置或 ?storage= 决定。
每个模型的结构相同图像、视频和音频均包含 type、url、content_type、width、height 和 duration。output 是模型自身的结果对象。
输出字段
video
object {url, content_type, file_name, file_size}模型生成的视频。
expanded_prompt
string | null供模型使用的扩展提示词;如果没有可用的扩展文本,则为 null。以下情况会返回 null:禁用扩展、扩展未产生变化,以及扩展由模型托管的 API 内部处理。
timings
object | null以秒表示的计时数据,仅在报告后端计时的路由上可用;否则为 null。'inference' 条目表示后端 GPU 执行 DiT 去噪所用的时间。

更多模型:图生视频

对于 AI 代理:此页面的 Markdown 版本位于 https://doitong.com/zh/developers/models/minimax/h3-max/camera-controls.md. 所有模型和指南: https://doitong.com/developers/llms.txt