Prompt that actually lands
Subject, action, light, and timing stay in the brief. Independent boards have put H3 Max first for prompt understanding, not just overall look.
Type a shot or drop a still. MiniMax H3 Max returns a 5–15 second clip with synced stereo audio — often before a five-second preview has time to finish playing.
Brief the shot, pick duration and frame, then render. Speed is the point — quality still has to survive the cut.
Model
MiniMax H3 Max · 5–15s · 480p / 768p
YOUR IDEA. NOW MOVING.
H3 Max turns a written shot into a clip you can review in seconds.
Recent renders
No generations yet.
A working creative tool
H3 Max is fal’s post-trained MiniMax H3: stronger prompt following and aesthetics, then an inference stack built so a 5-second 768p clip can land in under 3 seconds.
Subject, action, light, and timing stay in the brief. Independent boards have put H3 Max first for prompt understanding, not just overall look.
Call a track, a slow push, a locked frame, or a hard cut. The model is used for shot language, not only pretty motion.
Start from a still when the composition is already right. Optionally aim the clip at an end frame so the move has a destination.
Keep a character or product readable while location, light, and camera distance change — useful when you need several takes, not one lucky still.
Palette, texture, and material hold together well enough that a second pass is a note, not a restart.
A 5-second 768p clip often returns in ~2.5–3 seconds — about 35× the official MiniMax H3 endpoint. That is what makes storyboard loops practical.
Designed for the next take
Video models used to force a wait. H3 Max crossed the line where generation can finish before playback does, which is why people are already using it for live direction, ads, and rapid previz.
Write the shot as subject + action + camera + light + time. Change one variable, render again, and keep the take that reads. Native stereo audio comes with the picture instead of as a later patch.
If a still already carries the product or character, switch to image-to-video. Keep the composition, add motion, and optionally steer toward an end frame before you spend a full production day.
Starting briefs
These are shot briefs, not finished films. Use them to see how H3 Max handles look, camera, identity, and timing in 5–15 seconds.
Keep palette, material, and linework as one system while the camera moves through the scene.
A moving subject plus a deliberate follow path — speed, framing, and camera intention in one clip.
A character-led prompt that tests identity across setting, light, and camera distance.
A controlled morph with a clear start and finish, so timing does the work instead of noise.
Open on a reference still and drive the composition toward a planned end pose.
A compact performance: framing, expression, and scene sound in one short render.
Made for daily workflows
Use a short render to make the idea visible, compare variations, and give the next decision a moving image instead of a paragraph.
Test a hero move, a pack close-up, or a launch hook before the campaign is locked.
Vertical, square, or wide clips for feeds and launches — fast enough to try three versions in a coffee break.
Turn a board frame or written beat into motion for treatments, edits, and early reviews.
Explore a repeatable subject, a look, or a performance across a short sequence.
Transitions, kinetic layouts, and rhythm while the concept is still cheap to change.
Start from an image that already works and add motion without throwing away the composition.
Three focused steps
Say what changes, how it is seen, and what the audio should carry. Camera and light belong in the prompt.
Choose text or image, 5 / 10 / 15 seconds, and 480p or 768p. Add a start or end frame when the landing matters.
Review, download a useful take, then change one variable. The model is fast enough that iteration is the workflow.
Pay for generations, not a waiting room. Packs and plans cover 5–15 second clips; longer and higher-res takes cost more credits.
Pro
Best value — more credits each year.
Max
Maximum yearly volume with commercial rights.
Output specifications
5–15s
Clip length per generation, with native stereo audio
480p / 768p
Current H3 Max output. Built for speed, not a 2K beauty pass
1080p / 2K / 4K
Not on H3 Max today — use original MiniMax H3 when you need higher res and heavier edit tools
H3 Max vs MiniMax H3
Same H3 family, different jobs. Max is the fast hosted lane for short clips. Original MiniMax H3 is the higher-res, more open, more controllable one.
| Dimension | H3 Max (fal post-trained) | MiniMax H3 (original) | Who wins? |
|---|---|---|---|
| Generation speed | Extremely fast — a 5-second clip usually finishes in under 3 seconds, about 35× the throughput | Noticeably slower | H3 Max, by a wide margin |
| Prompt following & aesthetics | Post-training lifts look and prompt adherence; it often ranks higher in human preference tests, especially image-to-video | Already strong. Max still edges it in several blind tests | H3 Max, slightly |
| Max resolution | 480p / 768p | Up to 2K | Original H3 |
| Feature surface | Mainly text-to-video and image-to-video, with start and end frames | Multimodal references (image / video / audio), editing, and heavier control | Original H3 is broader |
| Weights | Closed — hosted only | Open weights you can download and run locally | Original H3 is more open |
| Best used for | Fast iteration, batch shorts, anything that needs a clip back before playback finishes | High resolution, fine control, reference assets, and local deployment | Different jobs |
FAQ
Speed, modes, output limits, audio, and how this site is billed.
Open the playground, write one clear beat, and review a clip while the idea is still cheap to change.