Forge is available: build a website with AI from your RodiumAi account.

Try Forge
MiniMax

MiniMax Hailuo 2.3 Fast

Healthyvideo
Context
-
Max output
-
Speed
Fast
Quality
Basic
Released
2025-10-28
Input: / Output:
ImageVideo

About this model

Fast MiniMax Hailuo 2.3 model for short cinematic video. MiniMax Hailuo 2.3 Fast is the speed tier of the Hailuo 2.3 video family.

Pricing

RateRODIUSD (ref.)Unit
Video
Per second24.0~ 0.0316667per second (RODI/s, USD/s)
1080p with audio41.7~ 0.055per second (RODI/s, USD/s)
1080p no audio41.7~ 0.055per second (RODI/s, USD/s)

RODI prices include RodiumAi markup and upstream fees. USD figures are wholesale reference rates.

Providers & pricing

ProviderInputOutputCachedFree tierStatus
Runware

Runware

Default

All rates (3)
  • Per second24.0 RODI/s
  • 1080p with audio41.7 RODI/s
  • 1080p no audio41.7 RODI/s
----No data

Capabilities

Streaming
Tool calling
Vision
JSON mode
Reasoning

API usage

Generate video with client.videos(): POST /v1/videos/generations (long timeout).

Examples

…

Example response

…

Text only (no image)

…

RodiumAi SDK

videos() maps to POST /v1/videos/generations: allow up to 600s timeout.

…

Use cases

  • Bring a product photo, a poster or an illustration to life.
  • Animate AI-generated stills into short clips.
  • Keep a character or a scene consistent: start from the same first frame.
  • Turn a storyboard frame into a motion test.
  • Create looping hero banners from a brand visual.

Notes

  • Endpoint: POST https://api.rodiumai.io/v1/videos/generations, JSON body. model and prompt are required, image is the first frame.
  • image (also read from image_url, input_image) accepts a base64 data URL, raw base64 or {"b64_json": "...", "mime_type": "image/png"}. Models served by Google also accept a gs:// URI. Prefer inline base64: it works everywhere.
  • The output keeps the aspect ratio of the input image, so you usually do not need aspect_ratio or size.
  • A few models also take last_frame (end frame, frame interpolation). Check the model card.
  • Generation takes one to several minutes. Use a client timeout of at least 10 minutes and do not retry on timeout: a retry starts and bills a second video.
  • Billing is per second of video, or per generation for some models. See the pricing table above.
View all