MiniMax/minimax-h3/Released Jul 2026

MiniMax H3

Open weights omni-modal video with audio

Fine-tunableCommercial use

Video

About MiniMax H3

MiniMax H3 (Hailuo 03) is an omni-modal video generation model that handles text-to-video, image-to-video with optional first and last frames, and reference-guided generation from images, video clips, and audio in a single model.

It generates 4 to 15 second clips at 768P or 2K resolution with native audio. Reference inputs allow up to 9 images, 3 video clips (15 seconds total), and 3 audio clips for guiding subjects, motion, and voice timbre. First/last frame inputs and reference inputs are mutually exclusive within one request.

See what this model can do

A handful of real prompts and their outputs.

Nature Spectacle

Glacier Calving

"Large-scale physics and a matching thunderous crack, roar of ice and rushing water at 2K."

Dance Performance

Courtyard Flamenco

"Fast footwork and a full skirt turn with heel stomps, castanets and guitar on the soundtrack."

Aerial Cinematic

Highland Viaduct

"Smooth drone tracking, billowing steam and golden light, with chuffing and a whistle echoing across the valley."

Wildlife Macro

Hummingbird Feeding

"Macro slow motion with blurred wingbeats, iridescent feathers and the soft hum of wings."

Street Music

Bucket Drummer

"A crowded scene with reactions, a passing train and punchy drumming that echoes off the tiles."

Image to Video

Map Unrolling

"Image-to-video from a still first frame, with a smooth unroll, crane-up camera and parchment sound effects."

Action Sports

Cliff Dive

"A single tracking tilt that follows the full fall, with a splash and cheering crowd on the soundtrack."

Wildlife Spectacle

Starling Murmuration

"Huge numbers of birds moving as one flock at 2K, with the rush of wings in the audio."

Live Performance

Taiko Ensemble

"Synchronized group motion under stage lighting, with booming drums and shouts in the generated audio."

Ready to build with MiniMax H3?

Try MiniMax H3 in the Workbench to prompt it, compare outputs, and iterate on prompts without writing any code. When you're ready to ship, call the same model from our API and build your own apps on top of it.

curl -sSf -X POST https://hub.oxen.ai/api/ai/videos/generate \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $OXEN_API_KEY" \
    -d '{
  "model": "minimax-h3",
  "prompt": "A cat sitting on a chair",
  "aspect_ratio": "adaptive",
  "duration": 5,
  "resolution": "768P"
}'

Request parameters

Every field MiniMax H3 accepts in the request body. See the API reference for response formats.

POSThttps://hub.oxen.ai/api/ai/videos/generate

Field	Type	Required	Default	Description
`prompt`	`string`	Yes	—	Describe the video content, motion, and cinematography.
`input_image`	`string` uri · nullable	No	—	Optional first frame. Cannot be combined with reference images, videos, or audio.
`tail_image_url`	`string` uri	No	—	Optional last frame. Cannot be combined with reference images, videos, or audio.
`input_images`	`array<string>` nullable	No	—	Optional reference images for subjects or style, up to 9. Cannot be combined with first or last frames.
`input_videos`	`array<string>` nullable	No	—	Optional reference video clips for motion or style, up to 3 clips of 2 to 15 seconds (15 seconds total). Cannot be combined with first or last frames.
`input_audios`	`array<string>` nullable	No	—	Optional reference audio for voice timbre, up to 3 clips of 2 to 15 seconds (15 seconds total). Cannot be combined with first or last frames.
`aspect_ratio`	`string`	No	`"adaptive"`	`adaptive21:916:94:31:13:49:16`
`duration`	`integer`	No	`5`	Range: 4 – 15
`resolution`	`string`	No	`"768P"`	768P on self-hosted and fine-tuned deployments; 2K is available on the hosted API only. `768P2K`

Model JSON Video generation reference Async queue Models endpoint