Skip to main content
POST
Hailuo Series
MiniMax Hailuo video generation series. Two tiersminimax-hailuo-2.3 (the 2.3 main release, adds 15 camera-movement tags) and minimax-hailuo-2.3-fast (the 2.3 accelerated tier, faster and cheaper, ideal for previews / batch). Supports text-to-video (T2V) and image-to-video (I2V).

Pricing

Billed by resolution × duration, in $ / second.

Examples

After submission, poll status with GET /v1/tasks/{task_id}; see Task System for details.

Available Models

Common Parameters

string
required
Model ID. Allowed values:
  • minimax-hailuo-2.3
  • minimax-hailuo-2.3-fast
string
default:"generate"
Operation type, usually no need to set explicitly (the backend auto-routes by media fields). Allowed values:
  • generate — text-to-video (T2V)
  • image2video — image-to-video (I2V); pair with image_urls
string
required
Video description text, up to 2000 characters. Embed [tag] brackets in the text to trigger the corresponding camera movement (see Camera-movement tags below)
string
default:"16:9"
Frame aspect ratio. Allowed values:
  • 16:9 — landscape widescreen
  • 9:16 — portrait
  • 1:1 — square
string
default:"768p"
Output resolution; see Available Models for per-model support. Allowed values:
  • 768p
  • 1080p
integer
Video duration in seconds. Allowed: 6 / 10 (1080p only at 6)
string[]
Reference image URL array; triggers I2V mode. Public URL or Base64
string
Webhook callback URL, invoked when the task reaches a terminal state. See Callback Mechanism
string[]
Subscribed callback event types; defaults to all terminal events

Model-specific Parameters

boolean
default:"false"
Whether to overlay a MiniMax watermark on the output video
boolean
default:"true"
Auto-optimize the prompt; when disabled, generation runs against the original prompt
boolean
default:"false"
Shortens prompt-optimization latency
string[]
Camera-movement tag list (15 total). Recommended to also embed [tag] brackets directly in the prompt; see Camera-movement tags below

Camera-movement Tags (2.3 / 2.3-fast)

The signature feature of the Hailuo 2.3 series — embed [tag] brackets in the prompt to trigger the corresponding camera movement; multiple tags can stack in a single sentence and the backend parses them in order. You can also pass them explicitly via the camera_movement_tags field for higher recognition reliability.
Example"[推进] cobblestone street of an old town, [变焦推近] raindrops falling from the eaves" — camera pushes forward first, then focuses on the eaves detail.

Resource Limits