MiniMax Hailuo

What is MiniMax Hailuo?#

MiniMax Hailuo is MiniMax's video generation lineup, and MiniMax H3, released July 31, 2026, is its current flagship. All four variants share the same /v1/tasks shape: minimax-h3 renders selectable 768p or 2K video at 24 fps with native audio, 5 to 15 second duration, and reference-to-video control; minimax-2.0 keeps start-and-end frame control at 768p or 1080p; minimax-2.3 auto-routes between text-to-video and image-to-video; minimax-2.3-fast runs the image-led path with shorter turnaround. Prompts run up to 2000 characters on every variant.

What's new in MiniMax H3#

MiniMax H3 supports selectable 768p and 2K output across 16:9, 9:16, 1:1, 4:3, 3:4, and 21:9 aspect ratios. Duration is any whole second from 5 to 15 instead of a 6-or-10 choice. Audio generates in the same pass as the video. And input control widens past frames: subject reference images lock a character across generations, reference videos drive continuation and instruction-based editing, and reference audio carries a sound identity forward. On the Artificial Analysis leaderboards (July 31, 2026), MiniMax H3 sits #1 in Video Editing (Elo 1,130), #2 in Text-to-Video with audio at Elo 1,242, behind Google's Gemini Omni Flash, and #3 in Image-to-Video with audio at Elo 1,185.

Key features of MiniMax Hailuo#

Best for#

Variants#

Hailuo has four models on the same API, and each has its own page with full details, playground, and FAQs.

minimax-h3#

The flagship. Selectable 768p or 2K output at 24 fps, native audio, 5 to 15 second duration, six aspect ratios, and three input styles: text alone, start/end frames, or references (up to 9 subject images, 3 videos, 3 audio files). A seed parameter makes runs reproducible. The prompt is required on this model.

minimax-2.0#

The start-and-end frame model. Pass a starting frame and, optionally, an ending frame, and the model animates the connective motion. Resolution options are 512p (I2V only), 768p, and 1080p.

minimax-2.3#

The auto-routing model. Pass a prompt, a start image, or both, and minimax-2.3 picks between text-to-video and image-to-video automatically. Useful when the upstream form lets users mix inputs and you do not want branching call sites.

minimax-2.3-fast#

The image-led speed model. start_image_url is required, the prompt is optional, and turnaround is shorter than 2.3. Use it when the workflow is image-led and the latency budget is tight.

Use cases#

Run a product campaign where the same mascot appears in ten videos by sending the character sheet as reference images to minimax-h3. Edit an existing video by passing it as a reference video with an instruction prompt: swap the product on the table, relight the scene, change the weather. Render a 21:9 teaser at 2944x1248 from a text prompt for a landing-page hero. Build a 6-second 1080p ad with a fixed open and close through minimax-2.0 with start and end frames. Keep one mixed-input call site on minimax-2.3, and route image-only bulk jobs through minimax-2.3-fast.

Limitations#

MiniMax H3 caps at 15 seconds per generation and 24 fps, with 768p and 2K output but no 4K tier; rivals like Kling 3.0 render native 4K. Reference media and frame images are mutually exclusive in one request. On the 2.x variants, 1080p locks duration to 6 seconds, and 10-second runs drop to 768p.

API examples#

Call MiniMax Hailuo from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/minimax/hailuo.

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "hailuo/minimax-h3",
    "input": {
      "prompt": "A slow dolly shot across a rain-soaked neon street, reflections rippling in the puddles",
      "aspect_ratio": "21:9",
      "resolution": "2K",
      "duration": 10
    }
  }'

Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished result.