视频生成

支持文生视频与图生视频。由于渲染需要数分钟，视频生成是一个异步任务：提交提示词后会收到一个 task_id，轮询至任务完成，然后下载 MP4。可用模型包括 grok-imagine-video、grok-imagine-video-1.5，以及多媒体分组的 Seedance / Wan 3.0 / Berry / MiniMax H3 等（见下文）—— 完整列表与价格请查看模型广场。

第 1 步 · 提交任务

curl https://cn.chatapi.app/v1/video/generations \
  -H "Authorization: Bearer sk-xxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine-video",
    "prompt": "A golden retriever puppy running on grass, 4 seconds"
  }'

import requests

base = "https://cn.chatapi.app"
key  = "sk-xxxxxxxx"
headers = {"Authorization": f"Bearer {key}"}

resp = requests.post(
    f"{base}/v1/video/generations",
    headers=headers,
    json={
        "model": "grok-imagine-video",
        "prompt": "A golden retriever puppy running on grass, 4 seconds",
    },
)
task_id = resp.json()["task_id"]
print(task_id)

响应会立即返回一个已排队的任务：

{
  "id": "task_UrufFh4Ww57Q6XWZ7trGZJajiHGw88Md",
  "task_id": "task_UrufFh4Ww57Q6XWZ7trGZJajiHGw88Md",
  "object": "video",
  "model": "grok-imagine-video",
  "status": "queued",
  "progress": 0
}

请求参数

第 2 步 · 轮询状态

建议每 5–10 秒轮询一次。status 会依次经过 queued → IN_PROGRESS → SUCCESS（或 FAILURE）。当变为 SUCCESS 时，响应中会携带 result_url。

curl https://cn.chatapi.app/v1/video/generations/task_UrufFh4Ww57Q6XWZ7trGZJajiHGw88Md \
  -H "Authorization: Bearer sk-xxxxxxxx"

完成后的响应（已省略部分字段）：

{
  "code": "success",
  "data": {
    "task_id": "task_UrufFh4Ww57Q6XWZ7trGZJajiHGw88Md",
    "status": "SUCCESS",
    "progress": "100%",
    "result_url": "https://cn.chatapi.app/v1/videos/task_UrufFh4Ww57Q6XWZ7trGZJajiHGw88Md/content"
  }
}

第 3 步 · 下载视频

请求 result_url（或根据 task_id 拼出内容地址），即可下载渲染完成的 video/mp4。

curl -L https://cn.chatapi.app/v1/videos/task_UrufFh4Ww57Q6XWZ7trGZJajiHGw88Md/content \
  -H "Authorization: Bearer sk-xxxxxxxx" \
  -o output.mp4

完整 Python 示例

import time, requests

base = "https://cn.chatapi.app"
headers = {"Authorization": "Bearer sk-xxxxxxxx"}

# 1. 提交任务
task_id = requests.post(
    f"{base}/v1/video/generations", headers=headers,
    json={"model": "grok-imagine-video",
          "prompt": "A puppy running on grass, 4 seconds"},
).json()["task_id"]

# 2. 轮询状态
while True:
    data = requests.get(f"{base}/v1/video/generations/{task_id}", headers=headers).json()["data"]
    if data["status"] == "SUCCESS":
        break
    if data["status"] == "FAILURE":
        raise RuntimeError(data.get("fail_reason"))
    time.sleep(8)

# 3. 下载视频
mp4 = requests.get(data["result_url"], headers=headers).content
open("output.mp4", "wb").write(mp4)

多媒体视频模型（未审查）

第三方未审查号池（支持 NSFW、严禁未成年内容），「按厂家」路由选「多媒体」或「全部模型与厂家」的 Key 均可调用。走 OpenAI Video API：POST /v1/videos 提交 → GET /v1/videos/{id} 轮询到 completed → GET /v1/videos/{id}/content 下载 MP4（成片保留 7 天）。一般 1–5 分钟。按秒计费，价格随分辨率变化，低于单次最低价按最低价收。同一系列的不同模式是不同的模型名（如 seedance-2.0-text-to-video / seedance-2.0-image-to-video）。

curl https://cn.chatapi.app/v1/videos \
  -H "Authorization: Bearer sk-xxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{"model":"seedance-2.0-mini-text-to-video","prompt":"a paper boat drifting on a rainy street","seconds":"5","size":"480p","metadata":{"ratio":"16:9"}}'
# → {"id":"task_xxx","status":"queued",...}；轮询 GET /v1/videos/task_xxx，completed 后下载 /v1/videos/task_xxx/content