MiniMax H3
Sign inStart free
Subscriber Perk

Get 50% Off MiniMax H3 and Other Selected Video Models

Unlock Savings

MiniMax H3 prompts, printed in full

Every prompt complete — no truncation, no paraphrase. Free to copy. 20 templates

Product

6
  • A white canvas sneaker spins mid-air and is hit by a burst of black and neon green paint
    P—02One shot

    Sneaker paint burst

    A labelled, seven-key prompt instead of a paragraph. The most literal read H3 will give you.

    8s16:92K$1.04
  • A crimson perfume bottle under a warm spotlight in a smoky dark-circus set
    P—082 steps

    Dark-circus perfume teaser

    A timed luxury commercial where the negative prompt is doing as much work as the shot list.

    15s9:162K$1.95
  • A striped cinema popcorn box on a red background in a fast-cut CGI commercial
    P—092 steps

    Ten-cut popcorn ad

    Ten shots in eight seconds, each with an explicit timecode. The tightest schedule here.

    8s3:42K$1.04
  • A woman sits still at a desk while the furniture showroom around her changes
    P—102 steps

    Showroom room swap

    Four rooms swap around a subject who never moves. One gesture drives every transition.

    15s9:162K$1.95
  • Four dogs modelling knitted sweaters against colour-matched studio sets
    P—112 steps

    Pet knitwear lookbook

    One four-shot unit repeated across four subjects, each in a colour-matched set.

    15s9:162K$1.95
  • A frozen fashion model on a metal chair with matcha suspended in mid-air as the camera orbits
    P—19One shot

    Matcha fashion orbit

    A 360° orbit around a frozen model and a frozen matcha splash. Camera only.

    15s1:12K$1.95

UGC

2
  • A creator holds an oversized croissant outside a bakery and speaks to camera
    P—122 steps

    UGC oversized pastry

    Five shots, five spoken lines, and one rule that keeps every one of them in sync.

    15s9:162K$1.95
  • A creator sits at a cafe table on a sunny street with a peach mini fan
    P—132 steps

    UGC problem to solution

    Problem, product, relief. The rule that makes it work is keeping both in frame together.

    15s9:162K$1.95

Performance

4
  • A woman in headphones speaks softly into a podcast microphone at a desk
    P—04One shot

    Podcast whisper take

    A single static shot with one written line of dialogue, delivered in sync.

    8s16:92K$1.04
  • A rapper performs across four escalating scenes, from a sports car to a fifty-storey giant
    P—03One shot

    Four-scene street rap

    Four locations in fifteen seconds with one performer who stays recognisable across all of them.

    15s16:92K$1.95
  • A woman in a black bomber performs a rap-video sequence in a blue studio
    P—16One shot

    Studio character rap

    A rap-video shot list that opens by locking one face and one outfit, then never lets them go.

    15s16:92K$1.95
  • Two streetwear kids under Tokyo neon, shot on a raw fisheye
    P—20One shot

    Tokyo fisheye streetwear

    A 13-second Tokyo night montage stitched by a fisheye and two lens-wipes.

    13s16:92K$1.69

Cinematic

6
  • Coffee and breakfast frozen in mid-air in a 1950s diner as the camera orbits
    P—06One shot

    Diner time freeze

    A thirty-second timed shot list cut down to fifteen — and what has to go.

    15s16:92K$1.95
  • A retro film camera held under a streetlamp on a wet black and white city street
    P—05One shot

    Midnight frame

    Monochrome product film where the mechanical sound design carries the whole shot.

    8s1:12K$1.04
  • An astronaut floats inside a derelict space station lit red by Mars as the camera spirals
    P—072 steps

    Mars station spiral

    A painted frame plus three sentences of camera. Everything else is already in the image.

    10s16:92K$1.30
  • A woman beside a zero-gravity pool in an orbital lounge with Earth through the window
    P—152 steps

    Zero-g party

    Eleven shots in ten seconds with no timecodes. Consistency comes from a lens and a palette.

    10s9:162K$1.30
  • An elf princess in silver fights through a crowd of green-skinned orcs in daylight
    P—17One shot

    Elf orc pursuit

    A one-take elf-versus-orc prompt that counts every contact. Written for 30s; H3 will give you 15.

    15s16:92K$1.95
  • A cybernetic prisoner fights through a tumbling zero-gravity cell block
    P—18One shot

    Zero-g prison

    A 15-second prison fight in five timed blocks, held together by gravity that will not stay put.

    15s16:92K$1.95

Anime

2
  • Two anime fighters clash in an abandoned factory as sparks and crates scatter
    P—01One shot

    Anime factory fight

    The shortest prompt here. It buys motion by describing impacts rather than people.

    8s16:92K$1.04
  • A pink-haired anime mage fires a beam into a black dragon over a rain-soaked magic city
    P—142 steps

    Anime dragon duel

    The longest prompt here. Two references, and a written rule for which one wins a conflict.

    15s16:92K$1.95

Every clip on this page is a reference render published elsewhere, made with some other model. The prompts are what this library publishes; the clips are here so you can see an idea before paying to run it. Each one links to its source.

Two shapes

Some of these need a picture first. Eleven do not.

Every template is one of two shapes. It is the only distinction that changes what you have to do before you can generate anything.

Two step — 9 templates

A still, then the clip

Any image modelMiniMax H3

Product shots, people, characters, anything that has to look like a specific thing. Step one is an image prompt we wrote by reading the actual reference frame; paste it into Nano Banana, GPT Image, Flux, Qwen-Image, Seedream — whichever you already pay for. Step two hands that frame to H3. If you already own the photograph, skip step one entirely.

We do not run an image model here yet. When we do, the copy button on step one becomes a generate button and nothing else on the page moves.

One shot — 11 templates

Straight to the clip

MiniMax H3 — picture and sound in one pass

Nothing to upload and nothing to prepare. H3 writes picture and sound in the same pass, so a sentence about a wet street and a shutter click comes back as both. Every workflow built on a silent video model needs an image pass and then a soundtrack; these need one prompt.

Copy the text, set duration and ratio, generate.

Before you copy anything

MiniMax H3 prompts FAQ

Ten answers · all visible · nothing collapsed

Are these MiniMax H3 prompts free to copy?

Yes. Every prompt on this site is printed in full and free to take, edit and use commercially. No account, no email, and copying never spends credits.

Are the prompts complete, or truncated?

Complete. Each one shows its character count on the same rule as the text, so you can check what you pasted against what we published. That number is computed from the string, not typed in.

Were these clips generated with MiniMax H3?

No, and every card says so. They are reference renders published elsewhere with other models, kept here so you can see what an idea looks like before spending on it. We are re-shooting the library on H3; when a clip is ours the label changes.

Why do some templates have two prompts?

Because nine of them start from a still — a product, a person, a character. The first prompt makes that frame in any image model, the second turns it into a clip on H3. The other eleven go straight to video from text.

Which image model should I use for step one?

Whichever you already have. The image prompts are written in plain structured English with no model-specific syntax, so they work in Nano Banana, GPT Image, Flux, Qwen-Image or Seedream. We do not run an image model here yet.

Why do some copied prompts come back wrong?

Usually length. H3 generates whole seconds from four to fifteen, and a prompt written for a thirty-second model is not truncated — the whole schedule is compressed into the runtime you asked for, so every beat lands early. Two templates here are dedicated to that edit.

Will the same prompt give me the same clip twice?

No. H3 exposes no seed, so every run is a different take. Treat the clip beside a prompt as one draw, not a promise.

How much dialogue fits in a clip?

About 2.5 words a second at a natural pace, so roughly twenty words in eight seconds. Past that the delivery is compressed rather than the clip extended, and lip sync is the first thing to go.

Where do I learn to write one from scratch?

This page exhibits finished examples. If you want the field structure explained and a validator to check your own text, that is the prompt generator.

What does running one of these cost?

It is printed on every card: runtime times the published per-second rate. Eight seconds at 2K is $1.04; fifteen seconds is $1.95. The arithmetic is visible so you can check it.

Take any of them. They are yours wherever you paste them.

Twenty templates, twenty-nine prompts, every character on the page.

Generate free