Skip to main content
Skip to main content

MiniMax H3

MiniMax H3 is MiniMax's open-weight video generation model - and the community reaction told the story faster than any spec sheet: within days of release, tutorials for running it on 8GB of VRAM were circulating alongside ComfyUI integrations on day zero. H3 generates video with synchronized stereo audio in a single pass, up to 15 seconds, at API output up to 2K.

  • 99.9% SLA
  • Official Discounts
  • Pay-as-you-go
  • Fast & Low Latency
Brand:MiniMaxPrice: From $0.0628 / sec

API DOCS / SIMPLIFIED

Integrate this model with one task-based API flow

MiniMax H3 is MiniMax's open-weight video generation model - and the community reaction told the story faster than any spec sheet: within days of release, tutorials for running it on 8GB of VRAM were circulating alongside ComfyUI integrations on day zero. H3 generates video with synchronized stereo audio in a single pass, up to 15 seconds, at API output up to 2K. The examples below are rendered from MiniMax H3, the current gateway base URL, and the selected public model contract.

Input mode
X-API-Key on every request
POST /api/v1/openapi/task
GET /api/v1/openapi/task/status/{task_id}
Async task workflow

Single code window

Switch language once. Each example already includes both the task creation request and the task status polling request.

Unified Example
API_KEY="<API_KEY>"
BASE_URL="https://api.musespark.ai"
REQUEST_BODY='{
  "model_id": "minimax-h3",
  "parameters": {
    "prompt": "A cinematic drone shot over a futuristic city at sunrise, volumetric light, realistic motion",
    "resolution": "2K",
    "duration": 5,
    "aspect_ratio": "21:9",
    "batch_size": 1
  }
}'

CREATE_RESPONSE=$(curl --request POST "$BASE_URL/api/v1/openapi/task" \
  --header "X-API-Key: $API_KEY" \
  --header "Content-Type: application/json" \
  --data-raw "$REQUEST_BODY")

TASK_ID="<TASK_ID_FROM_CREATE_RESPONSE>"

curl "$BASE_URL/api/v1/openapi/task/status/$TASK_ID" \
  --header "X-API-Key: $API_KEY"

Request and response JSON become the documentation

Keep the payload and task status envelope as the primary contract. Field descriptions, optionality, and enum hints now live inline as JSON comments.

Request Body

Caller payload

This is the request envelope accepted by the public gateway for the selected model contract.

Field comments and caller parameters come directly from the current item's model schema.
{
"model_id": "minimax-h3",// target model
"parameters": {
"prompt": "A cinematic drone shot over a futuristic city at sunrise, volumetric light, realistic motion",// prompt describing the video to generate · min 1 · max 7000 · min length 1 · max length 7000
"resolution": "2K",// output video resolution · enum: 2K | 768P
"duration": 5,// video duration in seconds. Must be between 4 and 15 seconds · min 4 · max 15
"aspect_ratio": "21:9",// output aspect ratio or adaptive sizing. Text-to-video defaults to 16:9; reference generation defaults to adaptive; frame images determine frame-generation dimensions · enum: 21:9 | 16:9 | 4:3 | 1:1 | 3:4 | 9:16 | Adaptive
"batch_size": 1// standard quantity control field. The allowed range for this item is 1-4 · min 1 · max 4
}
}
Response Body

Gateway response

This is the task status response you will query with task_id until the task reaches a terminal state.

Poll every 2-5 seconds and stop once status becomes completed, failed, cancelled.
{
"code": 200,// 200 on success
"msg": "success",
"data": {
"task_id": "task_01HQX8N2QJ0YQ8Q0T7G6N4V6K2",// use for status polling
"status": "completed",// enum: queued | running | completed | failed | cancelled
"progress": 100,// 0-100 if available
"completed_at": 1760000000,// unix timestamp
"expires_at": 1760001800,// unix timestamp
"result": {
"output_urls": [// generated files
"https://cdn.example.com/output.png"
]
}
}
}