MiniMax Hailuo 03 API: 1080p Video Generation

If you want 1080p AI video by API, the main limits Hailuo 03 brings are simple: 5-second max clips and async job handling. I’d treat Hailuo 03 as a short-form video model for apps that need text-to-video or image-to-video without running GPUs. One thing to know up front: APIMart does not offer Hailuo 03. The MiniMax video models available on APIMart today are MiniMax H3 and Hailuo 2.3, and this guide shows how to use them as alternatives.

Here’s the article in plain English:

• What it does: generates 1080p MP4 video
• Input types: text prompt, image-to-video, first-and-last-frame, and subject reference
• Clip limit: 5 seconds at 1080p, according to MiniMax
• Price: APIMart does not list Hailuo 03 pricing; MiniMax H3 and Hailuo 2.3 are billed by usage, with prices shown on their model pages
• API flow: submit job, then poll task_id or use callback_url (per MiniMax’s own API)
• Prompt control: bracketed camera moves like [Pan left] or [Zoom in]
• File handling: final video URL expires after 24 hours
• Image rules: under 20 MB and aspect ratio between 2:5 and 5:2, per MiniMax’s Hailuo documentation
• Reliability note: article cites 99.9% uptime SLA

What matters most is this: you need backend logic, not just a prompt. That means handling async status checks, storing the MP4 right away, retrying on 429 and 5xx, and stitching clips if you need anything longer than 5 seconds.

If I were setting this up, I’d test prompts at lower resolution first, lock the motion wording, then move to 1080p only for final runs to keep spend under control.

Key Takeaways

• According to MiniMax, Hailuo 03 generates 1080p MP4 video from text-to-video, image-to-video, first-and-last-frame, and subject-reference inputs, with 1080p clips capped at 5 seconds.
• Hailuo 03 is not available on APIMart. Available alternatives are MiniMax H3 (native 2K or 768P, 4–15 seconds, synced audio) and Hailuo 2.3 (6–10 second clips, default 768p).
• On APIMart, send a POST request to https://api.apimart.ai/v1/videos/generations with a Bearer token and model set to MiniMax-H3, then poll GET /v1/tasks/{task_id} every 5–10 seconds until status is completed.
• APIMart does not publish Hailuo 03 pricing. MiniMax H3 and Hailuo 2.3 are billed by usage at the prices shown on their model pages; test prompts with short clips at lower resolution before your final runs to estimate budget.
• Download the final MP4 right away, retry on 429 and 5xx with exponential backoff, and choose clip length based on the model: H3 supports 4–15 seconds, Hailuo 2.3 supports 6–10 seconds.

Watch: Automating MiniMax Video Generation

Core Capabilities and 1080p Output Options

Before you send your first request, get clear on Hailuo 03's input modes, motion controls, and output limits.

Supported Inputs: Text Prompts, Images, and Motion Instructions

According to MiniMax, Hailuo 03 supports four input modes: text-to-video, image-to-video (I2V), first-and-last-frame video, and subject-reference video ^[2].

For motion control, you can combine up to three camera moves inside one bracketed instruction, such as [Pan left, Pedestal up] ^[3]. That gives you a simple way to guide framing and scene movement without adding extra metadata.

If you’re working on APIMart, MiniMax H3 covers similar ground with its own modes: text-to-video, first-frame / last-frame / first-and-last-frame image-to-video, and multimodal reference-to-video.

1080p Output Specs Developers Should Verify

Per MiniMax, Hailuo 03’s 1080p output is capped at 5-second clips. If you need a longer sequence, generate multiple clips and stitch them together in your backend. For projects requiring integrated audio, consider Google's Veo 3.1 as an alternative; MiniMax H3 on APIMart also outputs a synced audio track.

That limit should shape both your request settings and your backend assembly logic.

Hailuo 03 Specs

MiniMax Hailuo 03 on APIMart: Current Status and Alternatives

Now that you’ve seen what Hailuo 03 can do, here is the current status: APIMart does not offer Hailuo 03, and there is no Hailuo 03 model ID on APIMart. If you want to build MiniMax video generation into your app through APIMart, use MiniMax H3 or Hailuo 2.3. They are separate models, not Hailuo 03, but they cover the same text-to-video and image-to-video workflows. The steps below use MiniMax H3; Hailuo 2.3 (MiniMax-Hailuo-2.3 or the faster MiniMax-Hailuo-2.3-Fast) uses the same submit and query endpoints.

Authentication, Base URL, and Headers

Every request to APIMart uses a Bearer token in the Authorization header, along with Content-Type: application/json. A single APIMart API key handles every request.

POST https://api.apimart.ai/v1/videos/generations

Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

Request Examples for Text-to-Video and Image-to-Video

You’ll use the same endpoint and the same auth setup each time. What changes are the prompt, resolution, duration, and input URL.

The examples below call MiniMax H3, the alternative model available on APIMart; they are not Hailuo 03 requests. The model field is MiniMax-H3 (or MiniMax-H3-Max). prompt is required and can be up to 7,000 characters. resolution is 2K (default) or 768P, and duration is an integer from 4 to 15 seconds (default 5).

Text-to-video request (MiniMax H3):

{
  "model": "MiniMax-H3",
  "prompt": "A product designer sketching at a sunlit desk, the camera pans left and slowly zooms in, cinematic depth of field",
  "resolution": "2K",
  "duration": 5,
  "aspect_ratio": "16:9"
}

Image-to-video request (MiniMax H3):

{
  "model": "MiniMax-H3",
  "prompt": "The product rotates slowly on a white surface as the camera orbits to the right",
  "resolution": "2K",
  "duration": 5,
  "first_frame_image": "https://your-storage.com/product-shot.jpg"
}

For image-to-video with H3, upload the image first and pass its URL in first_frame_image. Add last_frame_image as well if you want first-and-last-frame generation. When you don’t pass an image, H3 runs text-to-video. For Hailuo 03, MiniMax documents a 20 MB size limit and an aspect ratio between 2:5 and 5:2; don’t assume the same limits apply to H3, and check the model page before going to production.

Async Responses, Job Status, and Final Video URLs

Video generation runs asynchronously, so your app needs to poll for status. A successful submission returns a task ID:

{"code":200,"data":[{"status":"submitted","task_id":"task_..."}]}

"After submitting a task, poll its status using the task_id until it succeeds or fails." - MiniMax API Docs ^[2]

Poll this endpoint every 5–10 seconds (2K generation with H3 takes a while):

GET https://api.apimart.ai/v1/tasks/{task_id}

The status field moves through a few stages:

When the status reaches completed, the video URL is in result.videos[0].url. Download the MP4 right away; result links like this can expire, for example after 24 hours ^[4].

MiniMax’s own API also lets you pass a callback_url for Hailuo jobs instead of polling. Your server gets a POST callback when the job finishes, and it must return the challenge value within 3 seconds ^[3]. The H3 examples above don’t use a callback; they rely on polling GET /v1/tasks/{task_id}.

With the job flow set up, the next move is dialing in quality and cost for production. You might also consider Kling V3 for alternative cinematic video styles.

Parameters, Performance, and Pricing for 1080P Workloads

Quality Controls That Matter in Production

Once your request format is set, the next step is tuning output quality, speed, and spend. With the MiniMax H3 example on APIMart, the main levers are resolution, duration, aspect_ratio, and the prompt itself.

In MiniMax’s Hailuo API, prompt_optimizer rewrites prompts to make motion and composition clearer ^[1][3], and you can set it to false when your prompt needs to stick closely to brand terms or exact wording ^[3]. The H3 examples in this guide don’t use this parameter, so write the prompt exactly the way you want it.

The same applies to fast_pretreatment, which MiniMax describes as a way to cut prompt prep time at the cost of a small drop in output quality ^[1][3]. It isn’t part of the H3 examples here.

For camera movement in Hailuo 03, MiniMax supports bracketed commands in the prompt, such as [Pan left] and [Zoom in], with up to three commands in a single prompt ^[3][5]. With H3, describe camera motion in plain language inside the prompt.

Latency and Cost Planning in USD

After those controls are in place, cost mostly comes down to clip length and resolution. Since generation runs asynchronously, plan for a submit-and-poll flow ^[4].

APIMart does not offer Hailuo 03, so there is no Hailuo 03 price on APIMart. MiniMax H3 and Hailuo 2.3 are billed by usage; check the current prices on the MiniMax H3 and Hailuo 2.3 model pages, then run a small batch of short clips to estimate your budget.

One simple way to cut waste is to test prompts at a lower resolution first (768P with H3), then switch to the final resolution once prompt behavior and camera motion look right ^[1][6].

Integration Patterns and Next Steps

Backend Workflow for Marketing, Product, and Education Apps

With request handling and job status set up, the next move is putting your video model (MiniMax H3 or Hailuo 2.3 on APIMart) into actual product flows. The core job flow stays the same across app types. What changes is the prompt style, the input you send, and what the clip needs to do.

For marketing ad clips, use text-to-video. Keep prompts short and direct, and include clear camera cues such as a left pan or a tracking shot. For product visuals, use image-to-video and pass product shots as the first-frame image. For educational explainers, 768P is often the practical pick when you need longer clips.

Storage, Delivery, and Usage Tracking at Scale

Once rendering finishes, move the file into persistent storage for delivery and tracking. Download each MP4 right away and store it in your own system for delivery. For reliability, add exponential backoff on 429 and 5xx responses. If you're handling high volume, keep polling intervals reasonable and spread out submissions. Track usage in one place across all video jobs. That setup helps keep delivery steady as volume grows.

Conclusion: Key Points for Developers

Hailuo 03 is useful to understand, but on APIMart you build with MiniMax H3 or Hailuo 2.3. Submit jobs with the right input type for the use case, handle the async flow with care, and store output right away - then build from there.

FAQs

Is Hailuo 03 available on APIMart?

No. APIMart does not offer Hailuo 03 and does not list a Hailuo 03 price. The MiniMax video models available on APIMart are MiniMax H3 and Hailuo 2.3. They are different models from Hailuo 03, but you can use them as alternatives through the same async submit-and-poll API, with usage-based pricing shown on each model page.

How long does a 1080p video usually take to generate?

Reported generation times for high-quality 1080p Hailuo video usually range from 1 minute 38 seconds to 5 minutes, although some jobs wrap up in 30 to 90 seconds.

The exact timing comes down to two things: how complex your prompt is and how long you want the video to be. Since generation runs asynchronously, your app should poll the task status until it’s complete. With MiniMax H3 on APIMart, 2K generation is slower, so poll every 5–10 seconds.

What’s the best way to make videos longer than 5 seconds?

With Hailuo 03, MiniMax caps 1080p at short clips, so the usual approach is to use a lower resolution.

Per MiniMax, 1080p only supports 5- or 6-second clips, depending on the model version, while 768p supports clips up to 10 seconds.

On APIMart, you can also pick a model with longer clips:

• MiniMax H3: set duration anywhere from 4 to 15 seconds
• Hailuo 2.3: generates 6 to 10 second clips

For anything longer, generate several clips and stitch them in your backend.

When should I turn off prompt_optimizer?

prompt_optimizer is a parameter in MiniMax’s Hailuo API. Turning it off gives you tighter control, because by default the system rewrites your description to help improve the result. Switch it off there if you want your prompt used exactly as written.

The MiniMax H3 examples in this guide don’t use prompt_optimizer. With H3, the prompt you send is what drives the result, so state the subject, action, camera motion, and any brand terms explicitly, and refine the wording with short test clips before your final runs.