Sports and action clips
Generate motion-heavy footage where believable weight and momentum matter more than raw resolution.
Also known as Hailuo 3, Hailuo 03, MiniMax Hailuo 3
MiniMax H3 (Hailuo 3) - category-leading physics simulation, cinematic camera language, and synchronized audio in a single pass.
Best for Physics-heavy action scenes, cinematic B-roll with sound, and teams blocked from MiniMax direct access.
Access the MiniMax H3 video model via REST API on VideoGenAPI, with endpoints, a live playground and developer docs.
Explore the interface and see how easy it is to generate videos with MiniMax H3. Create an account and pick a plan to start generating.
Discover what's possible with MiniMax H3 AI video generation
Generated with the MiniMax H3 API at up to 1080p. Results vary with prompt detail and requested duration.
The MiniMax H3 API is a REST interface that lets developers generate AI videos programmatically using MiniMax's MiniMax H3 model. It supports text-to-video, image-to-video through a single POST endpoint. VideoGenAPI handles authentication, billing, and provider infrastructure - you only need one API key to start building with the MiniMax H3 API.
Integrate the MiniMax H3 API in your app in under 5 minutes with these four steps.
Sign up free at VideoGenAPI.com and generate your API key from the dashboard. No MiniMax account needed.
POST to https://videogenapi.com/api/v1/generate with your API key, model "minimax-h3", and a text prompt describing your video.
Use the returned generation_id to poll GET /api/v1/status/{id} until the job completes. Video completion webhooks are not available yet.
Once status is "completed", retrieve the video URL from the response and embed it in your app, website, or workflow.
MiniMax H3 is MiniMax's third-generation video model, also sold to consumers as Hailuo 3. MiniMax built its reputation on physics simulation, and H3 continues that line: falling objects accelerate correctly, cloth and hair respond to movement, and liquids behave believably instead of sliding across the frame. The model generates audio and video together rather than dubbing sound onto finished footage, so footsteps, impacts and ambience line up with what is on screen. Through VideoGenAPI you reach MiniMax H3 over the same REST endpoint as every other model, which matters because MiniMax restricts direct access in several regions including the US and EU. For side-by-side quality notes with Google and Kuaishou models, see Kling 3 vs Veo 3 vs Seedance (/blog/kling-3-vs-veo-3-vs-seedance).
Generate motion-heavy footage where believable weight and momentum matter more than raw resolution.
Produce clips that arrive with matching audio, skipping a separate sound design pass.
Show vehicles, machinery and moving parts behaving the way viewers expect them to.
Quick start guide to integrate MiniMax H3 in your application
| Parameter | Type | Required | MiniMax H3 values |
|---|---|---|---|
| model | string | Yes | "minimax-h3" |
| prompt | string | Yes | Scene description, up to 1,000 characters. |
| image_url | string | No | Public jpg, png, gif or webp URL. Animates the image (image-to-video). |
| duration | integer | No | 5 to 15 seconds, up to 15s on every plan. |
| resolution | string | No | 480p, 720p, 1080p. |
| aspect_ratio | string | No | 16:9, 4:3, 1:1 or 9:16 (text-to-video). |
| seed | integer | No | Fixed value for reproducible output, -1 for random. |
| add_audio | boolean | No | Adds an AI sound-effects track when set to true. |
Average generation time: ~88s. Poll GET /api/v1/status/{id} with the returned generation_id. Full reference: /docs/generate.
{
"model": "minimax-h3",
"prompt": "A skateboarder lands a trick in an empty concrete pool at dusk, board clatters on landing, dust kicks up, camera tracks low and close",
"duration": 10,
"resolution": "480p"
}
{
"success": true,
"generation_id": "gen_abc123xyz",
"status": "in_progress",
"estimated_time": 88}
const response = await fetch('https://videogenapi.com/api/v1/generate', {
method: 'POST',
headers: {
'Content-Type': 'application/json',
'Authorization': 'Bearer YOUR_API_KEY'
},
body: JSON.stringify({
model: 'minimax-h3',
prompt: 'A skateboarder lands a trick in an empty concrete pool at dusk, board clatters on landing, dust kicks up, camera tracks low and close',
duration: 10 })
});
const data = await response.json();
console.log(data.generation_id);
curl -X POST https://videogenapi.com/api/v1/generate \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "minimax-h3",
"prompt": "Slow push in, soft natural light, subtle movement",
"image_url": "https://example.com/your-image.jpg",
"duration": 5
}'
MiniMax H3 is billed per video, not per second of output. See the cost calculator for your volume.
per MiniMax H3 video on the Starter plan
| Plan | Monthly | Videos | Cost / video |
|---|---|---|---|
| Basic | $29 | 1,500 | $0.019 |
| Starter | $49 | 4,500 | $0.011 |
| Unlimited | $199 | No cap | Flat |
How the MiniMax H3 API compares on resolution, clip length, generation speed and price. All models below share one API key.
| Model | Provider | Max resolution | Clip length | Avg. speed | Price / video |
|---|---|---|---|---|---|
| MiniMax H3 API This page | MiniMax | 1080p | 5-15s | ~88s | Included, from $0.011 |
| Kling 3.0 API | Kuaishou | 1080p | 5-10s | ~250s | Included, from $0.011 |
| Seedance 2.5 API | ByteDance | 1080p | 5-30s | ~150s | Included, from $0.011 |
| Veo 3 Fast API | Google DeepMind | 1080p | 8s | ~60s | $0.45 |
Switching between these models means changing the minimax-h3
value in your request body. The endpoint, authentication and status polling stay the same.
Create a key and call the MiniMax H3 API in minutes, plus 15 other models on the same endpoint.
Free account · Plans from $29/mo · Cancel anytime