MiniMax H3 Videos API Integration Guide

POST https://api.acedata.cloud/minimax/videos provides text-to-video, image-to-video, and audio-guided video generation through one endpoint. The mode is inferred from the supplied media; no action parameter is required.

¶ Request parameters

Parameter	Type	Required	Default	Description
model	string	No	minimax-h3	Currently only minimax-h3 is supported
prompt	string	Conditional	—	Describes content, motion, camera, and style; required for text-only generation
image_urls	string[]	No	—	1–9 public HTTP(S) image URLs
audio_urls	string[]	No	—	1–3 public HTTP(S) audio URLs
ratio	string	No	16:9	16:9 or 9:16
duration	integer	No	4	An integer from 4 to 15 seconds
async	boolean	No	false	Return a task ID immediately when true
callback_url	string	No	—	Public HTTP(S) webhook; enables asynchronous execution

At least one of prompt, image_urls, or audio_urls is required. If audio_urls is present, the mode is audio-guided. Otherwise, image_urls selects image-to-video; prompt-only requests use text-to-video.

Public pricing is $0.25 per second. Failed tasks are not charged.

¶ Text to video

curl -X POST 'https://api.acedata.cloud/minimax/videos' \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "minimax-h3",
    "prompt": "A red fox running through a snowy forest at dawn, cinematic tracking shot",
    "ratio": "16:9",
    "duration": 6,
    "async": true
  }'

¶ Image to video

{
  "model": "minimax-h3",
  "prompt": "Gentle camera push-in, soft morning light",
  "image_urls": ["https://cdn.acedata.cloud/b1c82e4937.png"],
  "ratio": "9:16",
  "duration": 8,
  "async": true
}

Up to nine reference images are supported. For multiple images, explain the subject, style, or scene relationship to preserve in the prompt.

¶ Audio-guided video

{
  "model": "minimax-h3",
  "prompt": "A dancer moving naturally to the rhythm",
  "image_urls": ["https://cdn.acedata.cloud/b1c82e4937.png"],
  "audio_urls": ["https://cdn.acedata.cloud/6f7d62b18b.wav"],
  "ratio": "9:16",
  "duration": 8,
  "async": true
}

Up to three audio references are supported. Images and prompt are optional in this mode, but at least one audio URL is required.

¶ Asynchronous execution

Video generation usually takes several minutes, so async: true is recommended. The API immediately returns a task_id; use the MiniMax H3 Tasks API to retrieve the result.

Alternatively, provide callback_url. AceDataCloud POSTs the same final payload returned by synchronous execution when the task finishes.

¶ Successful result fields

Field	Description
success	Whether the request succeeded
task_id	AceDataCloud task ID
trace_id	Request trace ID
data[].id	Same public task ID as the top-level field
data[].model	Public model used
data[].mode	text_to_video, image_to_video, or audio_guided
data[].video_url	Video on the AceDataCloud CDN
data[].state	succeeded on success
data[].duration	Final billed duration in seconds
data[].ratio	Output aspect ratio

¶ Common errors

HTTP	Code	Meaning
400	bad_request	Missing/invalid parameters or media-count limit exceeded
403	forbidden	Content safety check failed
413	request_too_large	Request body is too large
429	too_many_requests	Too many tasks are in progress
500	api_error	Service is temporarily unavailable
504	timeout	This wait timed out; the task is not resubmitted