Hailuo 03 Is Here: MiniMax's Multimodal Video Model Lands on Apiframe

MiniMax just shipped Hailuo 03, and it's a bigger release than the version number suggests. Also known as MiniMax H3, this isn't simply a faster or sharper Hailuo. It's a different category of model: a general-purpose multimodal system that understands text, images, video, and audio together, and generates or edits any of them from a single unified pipeline. It's live on Apiframe today, behind the same API key and endpoint as the rest of the catalog.

What Is Hailuo 03?

Every Hailuo release before this one was, at its core, a specialized video generator: you gave it a prompt and maybe a reference image, and it gave you back a clip. Hailuo 03 moves past that. It interprets creative intent across mixed inputs, text, multiple images, video clips, and audio, and produces coherent output that respects all of them at once, rather than bolting separate generate, edit, and reference tools together.

In practice, that shows up as three things developers actually care about:

• Native multimodal generation: one prompt can draw on text, several reference images, video clips, and an audio track at the same time, and Hailuo 03 fuses them into a single coherent shot instead of treating each input as a separate pass.
• Precise multimodal editing: swap a character, replace a background, clone and transfer a voice, or add a synchronized second subject to an existing clip, touching only what you asked for and leaving the rest of the shot intact.
• Production-ready output: the model is tuned for commercial deliverables, including legible on-screen typography, brand-consistent product shots, UI/UX motion, game cinematics, and short-form ad content.

The Open-Weight Commitment

Most flagship video models ship as closed, API-only products. MiniMax has committed to releasing H3's weights, which is a meaningfully different bet. The weights aren't published yet, so today the model is API-only like everything else. But the commitment matters if you're weighing vendor risk before building a feature around a video model: it points at a future where the model is inspectable and portable rather than permanently locked behind one hosted endpoint, one pricing structure, and one set of usage terms.

Nothing about the API changes when the weights land. If you build on Hailuo 03 through Apiframe now, you keep the convenience of a hosted endpoint and inherit the option to self-host later.

What's New, Capability by Capability

Omni Reference

Instead of a single reference image, Hailuo 03 accepts up to 9 images, up to 3 video clips (2 to 15 seconds each, 15 seconds combined), and audio in one request, for up to 12 reference files in total. Audio has to be paired with at least one image or video rather than used on its own. The model picks the best aspect ratio automatically across your mixed inputs, or you can set one explicitly. This is what makes multi-shot brand campaigns, consistent-character storytelling, and style transfer from a mood board practical from a single call.

Native Audio, Every Time

Every Hailuo 03 generation includes native stereo audio by default: ambient sound, synced dialogue, or score, generated alongside the video rather than added in a separate step. That removes a post-production stage most teams currently handle outside the model.

First and Last Frame Control

Supply a start frame, an end frame, both, or neither. With no image input at all, Hailuo 03 runs as plain text-to-video.

Higher Resolution, Longer Clips

Output duration now runs from 4 to 15 seconds at any whole second, instead of Hailuo 2.3's choice between 6 and 10, at 24 fps. Launch output is native 1440p (2K), roughly 3.7 megapixels depending on aspect ratio. MiniMax has announced a cheaper 768p tier that isn't live yet, and we'll expose it the moment it is.

Instruction-Level Editing

Point it at an existing video and describe the change: replace the cat with a dog, add one person on the left walking in sync with the others, remove the green screen and composite in a fairy-tale forest. Hailuo 03 applies the edit to the targeted region and leaves motion, lighting, and pacing everywhere else untouched.

Specs at a Glance

Where This Actually Gets Used

MiniMax built Hailuo 03 with commercial production in mind, and it shows in the kind of prompts it handles well:

• Brand films and cinematic content: trailers, commercial-style spots, and premium brand films built from multi-image keyframe sequences.
• Product and e-commerce marketing: apparel try-ons, product feature visualizations, and campaign assets that stay on-brand across a whole shoot.
• Digital experiences and UI/UX: interactive game UI mockups, mobile app motion, and product landing renders with stable, legible typography.
• Character-consistent storytelling: vertical fashion content, K-drama-style dialogue scenes, and short-form narrative work that holds a character's face and wardrobe across shots.
• Editing existing footage: subject swaps, wardrobe replacement, synchronized character addition, and green-screen-to-environment composites on real source video.

How to Use Hailuo 03 on Apiframe

If you're already using any model on Apiframe, this needs zero new setup: same endpoint, same API key, same async job pattern. Just point model at "hailuo-03":

curl -X POST https://api.apiframe.ai/v2/videos/generate \
  -H "X-API-Key: afk_your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "a neon-lit night market street, slow dolly forward, steam rising from a food stall, ambient crowd noise and distant music",
    "model": "hailuo-03"
  }'

You get a job ID back immediately:

{
  "jobId": "b2c3d4e5-f6a7-8901-bcde-f23456789012",
  "status": "QUEUED"
}

Poll /v2/jobs/:id, or set a webhookUrl on the original request and let Apiframe notify your server when the clip is ready:

{
  "id": "b2c3d4e5-f6a7-8901-bcde-f23456789012",
  "status": "COMPLETED",
  "model": "hailuo-03",
  "result": {
    "videoUrl": "https://cdn2.apiframe.ai/videos/b2c3d4e5-f6a7-8901-bcde-f23456789012.mp4"
  }
}

No MiniMax account, no separate SDK, no new billing relationship. Hailuo 03 sits behind the same one-API, many-models setup as Kling, Veo, Seedance, Runway, and every other video model already on Apiframe.

Pricing

Hailuo 03 is billed per second on Apiframe's usual prepaid credits: 22 credits per second of output at 2K. Reference video is billed at the same rate on top, so a 6 second clip driven by a 4 second reference clip bills 10 seconds. The first 5 reference images are included, and each one after that adds 7 credits. Reference audio is free.

A plain 6 second clip therefore costs 132 credits. As always, you only pay for successful generations, with no separate invoice and no rate card to negotiate. See the model docs page for the full table.

Try It Today

Hailuo 03 is available now on Apiframe, alongside every other major video model, behind one endpoint and one API key. Get an API key and generate your first clip, or read the docs for full parameter details. Questions? Join our Discord.