MiniMax H3
MiniMax's H3 video model. Text to video, image to video from a first and optional last frame, and video guided by up to five reference images. Every clip has a generated soundtrack.
- Model id:
minimax/minimax-h3 - Type: Video, from MiniMax
- Modes: Text to video, Image to video, Reference to video
- Prompt: up to 7,000 characters
- Canceling: Free at any point. A running generation stops at its next status check.
Parameters
Set these in params. Anything left out takes the default, which is what the price estimate assumes.
| Parameter | Accepted values | Default |
|---|---|---|
duration_seconds | 5, 4, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 | 5 |
resolution | 768p, 2k | 768p |
aspect_ratio | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 | 16:9 |
num_outputs | 1 only | 1 |
output_format | mp4 | provider's |
References
Pass media in references, each with a role and an upload_id or HTTPS url (Uploads). When mode is left out it is inferred from the roles.
| Role | Used as | At most |
|---|---|---|
first_frame | First frame of the video | 1 |
last_frame | Last frame of the video | 1 |
reference_image | Subject or style reference | 5 |
Pricing
Billed per second of output video, varying by resolution. The full price list for your account is on the dashboard's billing page and in pricing from GET /v1/models/minimax%2Fminimax-h3. Each request reserves its maximum price and is charged for what was produced (Pricing and credits).
Examples
Text to video
import { RawFrame } from "@rawframe/sdk";
const rf = new RawFrame();
const generation = await rf.generations.createAndWait({
model: "minimax/minimax-h3",
prompt: "A paper boat drifting down a rain-soaked city gutter, macro lens, golden hour",
params: { duration_seconds: 5, aspect_ratio: "16:9" },
});
console.log(generation.outputs?.[0]?.url);Vertical clip
import { RawFrame } from "@rawframe/sdk";
const rf = new RawFrame();
const generation = await rf.generations.createAndWait({
model: "minimax/minimax-h3",
prompt: "A street drummer under a neon sign at night, slow push-in, crowd murmuring",
params: { duration_seconds: 10, aspect_ratio: "9:16" },
});
console.log(generation.outputs?.[0]?.url);Muse Image 1.0
Meta's Muse Image model from the Meta Model API. Text to image and image editing with up to ten input images. Flat price per image. No seed or mask.
MiniMax H3 Max
The faster variant of MiniMax H3, at 480p or 768p. Text to video, image to video from a first and optional last frame, and video guided by up to two reference images. Every clip has a generated soundtrack.