Hailuo 03 (MiniMax H3) API
by MiniMax
768p / 2K video with stereo sound.
Generate 768p or native 2K video with synchronized audio using MiniMax H3, through one unified API. Bring your own images, clips, and voice samples as references, and pay per second of output with async jobs and webhooks.
- 768p / 2K
- Native resolution
- 15s
- Max clip length
- Stereo
- Co-generated audio
- 10–16 credits/s
- Per-second rate
Playground
Try Hailuo 03 (MiniMax H3) right here.
Capabilities
What's new in Hailuo 03 (MiniMax H3).
768p or native 2K with audio in one pass
Every clip renders at 768p or 2K with stereo audio generated alongside the picture, so dialogue, effects, and ambience are already in sync instead of dubbed on afterwards. Upgrade a 768p draft to 2K without starting over.
Omni reference input
Feed the model up to nine images, three video clips, and three audio samples in a single request. It carries subject identity, setting, motion, and voice into a brand new scene.
First and last frame control
Set the opening and closing frames and let the model generate coherent motion between them, or supply just a first frame to animate a still.
Clips from 4 to 15 seconds
Pick any whole number of seconds in the range rather than choosing between two fixed lengths, and pay only for the duration you actually asked for.
Per-second pricing
Billing tracks the length of the clip, so short iterations stay cheap while you dial in a prompt and only the final render costs full price. See pricing for the current rate.
One API, async jobs, webhooks
Submit a generation, poll the job, or receive a webhook when the clip is ready, behind the same endpoint and API key as the rest of the Apiframe catalog. No separate MiniMax account to manage.
Showcase
See what Hailuo 03 (MiniMax H3) can create.
Real outputs generated with Hailuo 03 (MiniMax H3) on Apiframe, each with the prompt behind it.
A jazz drummer mid-solo in a smoky club, a man in his 40s with short salt-and-pepper hair, cream shirt with rolled sleeves, gold watch, vintage Gretsch kit. Brushes on the snare, warm tungsten light, slow push-in from the crowd, stereo room ambience with the kit panned naturally. Photoreal 2K, 35mm.
A woman alto saxophonist in an emerald silk dress, late 30s, short cropped hair, playing a ballad in a smoky jazz club. Warm tungsten practicals, brass catching the light, slow lateral dolly, stereo room with the horn slightly right of center. Photoreal 2K, 35mm.
Extreme close-up of jazz drum brushes on a snare, steam of the cymbal, gold watch on a wrist entering frame, warm tungsten, shallow depth of field. Stereo: brushes left, ride cymbal right, club murmur in the room. Photoreal 2K macro.
A jazz vocalist at a vintage chrome microphone, a woman in her 50s, silver bob, black turtleneck, singing a quiet standard in a smoky club. Warm tungsten, slow push-in, stereo: intimate vocal center, piano left, brushes right. Photoreal 2K, 35mm.
Vertical 9:16 from the crowd of a smoky jazz club: a drummer in a cream shirt mid-solo, brushes on the snare, warm tungsten, handheld phone-style, stereo room ambience. Photoreal 2K.
First and last frame: the empty jazz club stage brightens as a warm tungsten spotlight fades up on the vintage drum kit, golden haze filling the room, no people. Stereo: a ride cymbal swell and room tone.
Pricing
Simple per-second pricing.
10–16 credits per second of output ≈ $0.10–$0.16
The rate depends on the resolution and options you pick.
A 9s clip runs 90–144 credits (≈ $0.90–$1.44).
- You only pay for successful generations. Failed jobs are refunded automatically.
- One credit balance across every model on Apiframe, no per-model plans.
- Start with free credits. No subscription required.
- Volume discounts on larger plans.
Specs
At a glance.
Comparison
Hailuo 03 vs Seedance 2.5 vs Kling 3.0
Hailuo 03 against the other flagship video models on Apiframe. All one parameter swap apart.
| Feature | Hailuo 03 | Seedance 2.5 | Kling 3.0 |
|---|---|---|---|
| Max resolution | 768p / 2K | 720p | 4K mode |
| Max clip length | 15s | 30s in a single pass | 15s |
| Reference inputs | Images, videos & audio | Up to 50 (30 images, 10 videos, 10 audio) | Start & end frames |
| Native audio | Stereo, co-generated | Co-generated with lip-sync | Optional |
| Editing | Instruction-based edits | Region-level local edits | — |
| Price | 10–16 credits/s (≈$0.10–$0.16) | 13–35 credits/s | 11–26 credits/s |
| Best for | High-detail 2K footage with stereo sound | Long-form narrative with references | Multi-shot marketing video |
Specs reflect what each model supports on Apiframe today, from the live catalog.
Quickstart
How to call the Hailuo 03 (MiniMax H3) API.
Send one POST /v2/videos/generate request with your API key. The call returns a jobId you can poll, or pass a webhook_url and the result is pushed to you the moment it's ready.
Swap hailuo-03 for any other model and nothing else changes.
const response = await fetch("https://api.apiframe.ai/v2/videos/generate", {
method: "POST",
headers: {
"X-API-Key": "afk_your_api_key_here",
"Content-Type": "application/json",
},
body: JSON.stringify({
"prompt": "a cinematic sunrise over a futuristic cityscape, smooth camera push-in",
"model": "hailuo-03",
"hailuoParams": {
"resolution": "2K",
"image": "https://example.com/input.jpg",
"end_image": "https://example.com/input.jpg",
"base_video": "https://example.com/input.jpg"
}
}),
});
const { jobId } = await response.json();
console.log(jobId); Input schema
Every field Hailuo 03 (MiniMax H3) accepts, with types and defaults. The full reference lives in the docs.
| Field | Type | Description |
|---|---|---|
| prompt required | string | Text description of what to generate. |
| model required | string | The model identifier for this endpoint. Default:"hailuo-03" · "hailuo-03" |
| hailuoParams.resolution | string | Resolution Default:"2K" · "2K", "768p" |
| hailuoParams.image | string (URL) | Use a still as the first frame. |
| hailuoParams.end_image | string (URL) | Optional last frame to control the ending. |
| hailuoParams.base_video | string (URL) | A 768p Hailuo 03 clip to regenerate at 2K. Output is always 2K. |
| hailuoParams.reference_images | string[] (URLs) | First 5 are included; extra images are billed per image. |
| hailuoParams.reference_videos | string[] (URLs) | 15s total across all clips. Billed per second, like the output. |
| hailuoParams.reference_audios | string[] (URLs) | Voice timbre reference. Needs at least one reference image or video. |
Hailuo versions
Every Hailuo video model on Apiframe. Switch with one parameter.
Use cases
What teams build with Hailuo 03 (MiniMax H3).
High-detail hero footage
Native 2K output (not an upscale) holds fine texture in skin, fabric, and typography, which is what makes clips survive a big screen or a crop.
Sound-designed clips
Stereo audio is co-generated with the picture, so ambience and effects arrive spatially placed instead of a mono track you replace in post.
Reference-driven consistency
Feed reference images, videos, and audio in one request to pin a character, a look, or a sonic identity across a whole campaign of clips.
Instruction-based editing
Describe the change, swap the background, change the wardrobe, and Hailuo 03 edits the shot instead of forcing a full re-roll.
Reviews
9,800+ developers. One API.
Real Apiframe reviews from Trustpilot and G2.
Seamless experience, excellent support
Just created 6 videos and the experience has been seamless. Wanted to write a review for appreciation — the customer support has also been excellent. They helped me in some process I was stuck in.
Stable, easy image & video generation
I like that I can generate images and videos without complicated setup. It works stable and is easy to use even if you are not a big developer. It saves me time and makes the whole process much simpler.
Honestly the best AI API
They got every engine possible and it's great to have them all in one place.
Hailuo 03 (MiniMax H3) questions
What teams ask before shipping with Hailuo 03 (MiniMax H3). Everything else lives in the docs.
What is Hailuo 03?
Hailuo 03 is Apiframe's model id for MiniMax H3, a multimodal video model that generates 768p or native 2K clips with synchronized audio from text, frames, or reference media.
What inputs does it accept?
Three modes, one per request. A text prompt on its own; a first and optional last frame; or a set of reference images, videos, and audio clips. Frame inputs and reference inputs cannot be combined in the same call.
Does it really generate audio?
Yes. Audio is produced in the same pass as the video and muxed into the returned mp4, covering dialogue, sound effects, and ambience. You can also supply an audio reference to steer the voice.
What resolutions and durations are supported?
768p or native 2K, at any whole-second duration from 4 to 15 seconds. Default is 2K. Pass a 768p Hailuo 03 clip as base_video to regenerate it at 2K.
How is it billed?
Per second of output. Reference videos are billed at the same per-second rate on top, and reference images past the first five add a small flat charge each. You only pay for successful generations.
Is the model open weight?
MiniMax has committed to releasing H3 weights, but they are not published yet. Nothing about the API changes when they land, and the hosted model here stays the fastest way to run it.
How do I use it through the API?
POST to /v2/videos/generate with model "hailuo-03", then poll the job or receive a webhook when the video is ready.
Related models
All video modelsFlux 3
by Black Forest Labs
Black Forest Labs' multimodal model — up to 20s of video with synchronized audio from text, images, or a clip.
View modelVeo 3.1
by Google
Google's flagship cinematic video model with audio.
View modelSora 2 Pro
by OpenAI
Highest-quality Sora 2 tier with HD option.
View modelKling 3.0
by Kuaishou
Latest Kling — per-second pricing, optional audio.
View model9,800+
developers & companies
Official partner of
Power your next AI product with Apiframe.
Instant access to 70+ media models through a single API. Start free and scale when you're ready.