How much does MiniMax H3 cost?

Priced per second by resolution: 7 gold at 480p and 14 at 768p, on every route. A 5-second 480p clip is 35 gold, a 15-second 768p clip is 210. Reference images add nothing.

What is the difference between image-to-video and reference-to-video?

One image is treated as the opening frame and animated. Two or more are treated as references: subjects and props the model composes into a new scene. Upload a second image and H3 switches modes automatically.

How long can a MiniMax H3 clip be?

Any whole number of seconds from 5 to 15. Pricing is per second, so you only pay for the length you pick.

Can I choose the resolution?

Yes. H3 outputs 480p or 768p on every route. 480p is the cheap draft tier at 7 gold per second, 768p is the quality tier at 14.

MiniMax H3 AI Video Generator

MiniMax's unified video model, covering text-to-video, image-to-video, and reference-to-video with up to nine images, all under one name. Clips run any whole second from 5 to 15 at 480p or 768p, so you are not stuck choosing between 5 and 10.

How to Master MiniMax H3

Step 1
Start From a Frame You Already Like
Upload one image and H3 animates from it as the first frame. This is the most reliable way to keep a character on-model, because the look is fixed before any motion is generated.
Step 2
Add References for a Cast, Not a Frame
Upload two or more images and H3 switches to reference-to-video: the images become subjects and props it composes into a new scene, rather than a starting frame it animates. Use this to put several characters in one shot.
Step 3
Pick the Duration You Actually Need
Every whole second from 5 to 15 is available and pricing is per second, so a 7-second beat costs 7 seconds' worth. There is no reason to round up to 10.

Example Prompts

Example 1

Slow push in on a hooded figure at a crossroads shrine, mist rolling past, lantern flame guttering, overcast dusk

Example 2

A dragon banks hard over a burning coastal town, camera tracks alongside, embers streaming past the lens

Example 3

Two adventurers meet in a rain-soaked alley, one lowers a hood, handheld camera, neon-free medieval lighting

Example 4

Wide establishing shot of a mountain keep at dawn, clouds moving below the walls, banners snapping in wind

Example 5

A spellcaster traces a sigil in the air, light builds and discharges, camera holds steady, dark stone chamber

💡 Click the copy button to use these prompts in your own generations

Technical Specifications for MiniMax H3

Model TypeMiniMax unified video engine

ModesText-to-video, image-to-video, reference-to-video

Resolution480p or 768p on every route

Duration5 to 15 seconds, any whole second

Aspect Ratios21:9, 16:9, 4:3, 1:1, 3:4, 9:16, 9:21

Reference ImagesUp to 9

Gold CostPer second: 7 at 480p, 14 at 768p; reference images are free

Frequently Asked Questions for MiniMax H3

About MiniMax H3 for CharGen

MiniMax H3 is MiniMax's unified video model, offered on CharGen as a single entry covering three routes. Prompt alone and it generates from text. Add one image and it animates that image as an opening frame. Add more and it treats them as a cast of references to compose into a new scene. You pick the mode by what you upload, not by switching models.

Key Strengths

The nine-image reference ceiling is unusually generous, which makes H3 a good fit for party scenes and recurring NPCs where several characters need to stay recognisable. Per-second pricing across a 5 to 15 second range means you can cut a shot to the length the beat actually needs, and the 480p tier at 7 gold per second makes it one of the cheapest ways to draft video on CharGen.