01
Desert Car Revs
A desert car shot driven by a single reference image — engine, dust, and heat.
Create with this directionAI VIDEO MODEL
The Fast MiniMax H3: fal's post-trained MiniMax H3 Max turns a written brief or a first-frame image into a finished 768P clip in seconds — up to 15 seconds, with native stereo audio.
CREATIVE DIRECTIONS
Each direction combines a subject with choices about motion, camera, and rhythm — brief it in words or set the opening frame.
The videos below were created with the MiniMax H3 model family.
01
A desert car shot driven by a single reference image — engine, dust, and heat.
Create with this direction02
A second generated take of the courtyard dance brief with the same shot-by-shot choreography.
Create with this direction03
A talking-portrait beat animated from a single reference image.
Create with this direction04
An aging warrior-monk climbs sun-blasted monastery ruins under surveyor drones, in post-apocalyptic sci-fi silence.
Create with this directionMODEL GUIDE
MiniMax H3 Max is post-trained by fal Research on MiniMax H3 and co-optimized with fal's inference stack. Recent evaluations rank it first for overall quality, prompt adherence, and aesthetics — a 5-second clip renders in roughly 3 seconds, about 50× faster than the base model.
In Orelon that speed becomes an editing loop: write the direction, generate, and revise one variable at a time while the cost of a throwaway take stays near zero. Clips run 5–15 seconds at 768P with a native stereo audio track.
CAPABILITIES
PROMPT ANATOMY
A useful prompt gives each creative variable a job, then keeps the direction specific enough to revise.
WORKFLOWS
READY TO SEND
Watermark removal, generators and every model page — one click away.
MiniMax H3 Max is available for text-to-video and image-to-video: start from a written prompt, or attach one image as the opening frame and let the model carry motion, audio, and the ending.
On fal's custom inference stack, H3 Max renders a 5-second clip in roughly 3 seconds — about 50× faster than the base model — while recent evaluations rank it first for overall quality, prompt adherence, and aesthetics.
Available clip-length choices range from 5 to 15 seconds.
Output options are available through 768p. Credit use is shown in the video workspace before you submit a generation.
MiniMax H3 Max is post-trained by fal Research on the open-weights MiniMax H3 model. H3 Max Turbo is a faster, lower-priced sibling from the same family; H3 Max trades a little of that speed for stronger prompt adherence and aesthetics.
The hero and creative-direction videos on this page were created with the MiniMax H3 model family, and the wall examples lower on the page are curated MiniMax H3 references for planning your own scenes.
Revise one creative variable at a time, such as the subject, action, sequence order, camera, or audio note, then compare how that single change affects the direction — renders are fast enough to make iteration cheap.
MORE DIRECTIONS
This lower wall curates MiniMax H3 video inspiration for planning your MiniMax H3 Max scenes.
Plans for independent creators
No hidden fees • Cancel anytime • Unused paid credits roll over
Trial
Billed monthly
360 credits / month
Credits never expire
Starter
Billed monthly
810 credits / month
Credits never expire
Creator
PopularBilled monthly
3,600 credits / month
Credits never expire
Pro
Billed monthly
9,000 credits / month
Credits never expire
Free
10 signup credits