H3 Max

What it is

H3 Max turns a written prompt or a single still image into a 5–15 second video
with synchronized audio and directed camera motion. There is no timeline, no
node graph, and nothing to install — you type the shot you want, watch it come
back, and adjust.

It runs on MiniMax's H3 family: H3 Max, post-trained for prompt adherence,
for fast 480P and 768P work, and standard H3 when a piece needs to finish at
2K or 4K.

Who it's for. People who need short shots quickly and often — social and
performance marketers testing creative, editors and directors blocking out a
sequence before committing a crew, solo creators without a studio behind them,
and product teams that need a moving hero shot this afternoon rather than next
quarter. If you are producing a feature, this is a previz tool. If you are
producing a week of content, it may be the whole pipeline.

Why this rather than the others.

• The shot arrives with its audio. Most generators hand back a silent clip
  and leave you to score it. This one doesn't, which removes an entire pass from
  the process.
• You pay by the second, for what actually rendered. Pricing is per second
  of output and the exact cost appears on the button before you commit. A render
  that fails returns its credits automatically — you are never billed for a
  video you did not receive.
• Iterating is cheap on purpose. Draft at 480P until the shot is right, then
  re-run the one you kept at 768P. The price difference between exploring and
  finishing is the point of having both.
• You can direct where a shot lands. Give it a first frame and a last frame,
  and the motion between them is yours to specify rather than to hope for.
• Nothing to set up. Sign in and generate. New accounts start with free
  credits, and an account costs nothing to keep.

Features

Text to video. A prompt is enough. Describe subject, motion, lens, and mood;
H3 Max was post-trained specifically to follow what you wrote.

Image to video. Upload one image and it animates. Add a second image as the
end frame and the model works out the movement between them.

Audio, generated with the picture. Sound arrives synchronized to the shot,
not layered on afterward.

5 to 15 seconds per render. Long enough to be a real shot, short enough to
iterate on. Longer pieces are assembled from several renders.

480P to 4K. H3 Max covers 480P and 768P. Standard H3 adds 2K and 4K for
work that has to hold up on a large screen.

Per-second credit pricing. Cost scales with the length and resolution you
actually chose, shown before you run it. Failed renders are refunded without
you asking.

A library of everything you have made. Every generation is kept with its
prompt and settings, so a shot from last week is one click from being re-run
with one word changed.

English and Chinese throughout.

Use cases

Ad and social creative. Produce a dozen variants of the same concept in an
afternoon, run them, and put the budget behind whichever one performs. The unit
of work is a shot, not a shoot.

Previz and storyboards. Show a client or a crew what the sequence looks like
before anyone books a location. Moving boards land differently than a stack of
stills.

Product and e-commerce. Animate a product still into a hero shot — a slow
push in, a turn, a change of light — without a table-top studio.

Bringing stills to life. Archive photographs, illustrations, concept art, key
frames. One image in, motion out, with the end frame under your control.

Music, mood, and title work. Short atmospheric pieces where the sound and
the image are generated together, which is exactly the case that benefits from
audio arriving with the picture.

Pitches and prototypes. A concept explains itself faster as ten seconds of
video than as a paragraph, and at this price it is worth making one just to find
out whether the idea reads.