Skip to main content
POST
MiniMax H3 video generation
Use text_to_video without media, or first_frame with exactly one image. Turbo does not support video_multi_ref. Use one field spelling per media kind. Replace legacy 720p with explicit 768P. Completed jobs without usable artifacts remain retryable; failure/cancellation takes precedence over stale artifacts. Status reads do not own billing settlement. Pricing follows the current catalog; no fixed generation latency is guaranteed. Query Video Task · Pricing

Authorizations

Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

application/json
model
enum<string>
required
Available options:
minimax/h3-max-turbo
prompt
string
required
Required string length: 1 - 50000
operation
enum<string>

Must agree with media: text uses none; first_frame requires one image only.

Available options:
text_to_video,
first_frame
image_url
string<uri>

HTTP(S) first-frame image; do not combine with image_urls.

image_urls
string<uri>[]

Ordered images; one field spelling per media kind.

Maximum array length: 1
resolution
enum<string>
default:768P

Case/whitespace normalized. 720p is rejected; select 768P explicitly.

Available options:
480P,
768P,
1080P
duration
default:5
Available options:
5,
10
aspect_ratio
enum<string>
default:16:9
Available options:
16:9,
9:16,
1:1
prompt_optimizer
boolean
default:true
image_with_roles
object[]

First-frame role form. Do not combine with image_url/image_urls; not accepted by video_multi_ref.

Required array length: 1 element

Response

Queued video task: id, task_id, model, status=queued, created

id
string
task_id
string
model
string
object
enum<string>
Available options:
video
status
enum<string>
Available options:
queued
created
integer