Video Gen Now

MiniMax H3 video generator

MiniMax H3 is a 2K video model with unusually stable subject motion — it holds a person or an animal together through a shot instead of letting them come apart halfway. All three modes are here, with no API key and nothing to install.

Describe the shot: subject, action, camera move, light. Two or three sentences beat one long list of keywords.

130 credits

Rendering is paid in credits, and new accounts start at zero. Pricing

Made with this model

Every clip below was rendered here, on the settings shown under it. The caption is the exact prompt that produced it.

A lone red kayak cuts across a glassy alpine lake at dawn; mist lifts off the water and the mountains catch the first light. Slow aerial orbit, cinematic wide shot.

MiniMax H3 · 2K · 16:9 · 5s

The cat stretches, blinks awake and settles back down; dust motes drift slowly through the sunbeam. Gentle handheld drift.

MiniMax H3 · 2K · 16:9 · 5s

The coral origami crane unfolds its wings and lifts off a wooden desk, drifting past a window into the evening light. Smooth tracking shot.

MiniMax H3 · 2K · 16:9 · 5s

Running H3 here

  1. Pick your mode

    Text to video for a scene from nothing, image to video to animate a still you already have, reference to video to lock a character across shots.

  2. Set the duration

    Five to fifteen seconds. Everything renders at 2K, so there is no resolution ladder to think about.

  3. Render

    A flat 26 credits a second — 130 for a five-second clip — no matter which aspect ratio you pick. The cost is on screen before you submit, and refunded in full if the job fails.

Specs at a glance

ModelDurationResolutionNative audioReference imagesFrom
MiniMax H35–15 seconds2KNoUp to 9130 credits for 5 seconds

What H3 is good at

Single-subject motion. A person walking, an animal running, a dancer — H3 keeps limbs attached better than most.

Steady camera language. Slow pushes, follows and pans come back clean; whip pans and complex rigs less so.

Vertical work. 9:16 at 2K is genuinely usable for feeds without an upscale pass.

Reference consistency. Give it two or three angles of a character and the face carries across separate renders.

MiniMax H3 questions

Compare and switch

AI video generatorTurn a prompt, a photo or reference images into a finished clip. Seedance 2.0 and MiniMax H3 in one editor, MP4 export, pay per render in credits.Open this page
Text to videoType a description and get a finished clip. Text to video on Seedance 2.0 and MiniMax H3, up to 4K, native audio, cost shown before you render.Open this page
Image to videoUpload a photo, product shot or illustration and describe the motion. Image to video on Seedance 2.0 and MiniMax H3, up to 4K, with synchronized native audio.Open this page
Reference to videoUpload up to nine reference images and keep the same character, product or style across every shot. Reference to video on Seedance 2.0 and MiniMax H3.Open this page
Seedance 2.0Run ByteDance Seedance 2.0 in your browser: text, image and reference to video, 480p to 4K, 4 to 15 seconds, with synchronized native audio.Open this page
Pricing