Generate AI video faster than you can watch it
MiniMax H3 Max renders a five-second 768p clip with synchronized audio in about two seconds. That is shorter than the clip itself — so writing a prompt stops being a submission and starts being a conversation. Type something below and watch the timer.
What MiniMax H3 Max is
MiniMax H3 Max is a post-trained version of MiniMax H3, served by fal as two endpoints: Turbo, which renders a 5-second 768p clip in about 1.5 seconds, and the standard endpoint, which takes roughly 2.7 seconds for the same clip. Both produce 480P or 768P (1344×768 at 24fps in 16:9), 5 to 15 seconds, in six aspect ratios, and both generate a synchronized audio track on every render — there is no separate audio call and no toggle.
The ceiling is 768P. If you need 2K, that is still the standard MiniMax H3 model, which renders in minutes rather than seconds.
Why “faster than playback” changes how you work
When a render takes two minutes, you write one careful prompt and hope. When it takes two seconds, you write a rough one and fix it — five times in the span it used to take to check on the first attempt. Three things become practical that were not before:
Iterative prompting. Change two words, run it again, compare. The Try a variation box above keeps your last prompt loaded for exactly this.
Variations. Run the same idea four ways and pick, instead of committing to the first output you were shown.
Storyboard drafts. Rough in a whole sequence shot by shot in a couple of minutes, then re-render only the shots that survive at a higher resolution.
Specifications and price
| Resolution | 480P and 768P (1344×768 @ 24fps in 16:9) |
| Clip length | 5, 10 or 15 seconds |
| Aspect ratios | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 — image to video follows the input image |
| Audio | Synchronized audio on every render, always on |
| Modes | Text to video, image to video (optional end frame) |
| Upstream price | Turbo $0.04/s at 768P, standard $0.08/s — 75% off until September 7, 2026 |
| Clip length | Turbo · 768p | Turbo · 480p | MiniMax H3 Max · 768p |
|---|---|---|---|
| 5s | 20 credits | 13 credits | 40 credits |
| 10s | 40 credits | 25 credits | 80 credits |
| 15s | 60 credits | 38 credits | 120 credits |
Prices are in credits and are charged per render, not per month. Work out what a batch costs with the MiniMax H3 Max cost calculator, or see the credit packs.
MiniMax H3 Max vs MiniMax H3 vs Seedance 2.5
| Model | Render time | Resolution | Audio | 15s clip |
|---|---|---|---|---|
| MiniMax H3 Max Turbo | ~1.5s for a 5s clip | 480p · 768p | Always on | 60 credits |
| MiniMax H3 Max | ~2.7s for a 5s clip | 480p · 768p | Always on | 120 credits |
| MiniMax H3 | Minutes | 2K | No | 390 credits |
| Seedance 2.5 | Minutes | 480p · 720p | Optional | 694 credits |
Render times for the MiniMax H3 Max endpoints are the inference times fal publishes on its model pages. The other two are our own observations on this site — they are queue-based models measured in minutes, not seconds.
Frequently asked questions
Is MiniMax H3 Max free to use here?
Every account gets one free clip — 5 seconds at 768p with audio, no card. Sign-in is required, and it is a one-time trial rather than a daily allowance: it does not come back the next day. Beyond that first clip, and for 10 or 15 second clips, generation runs on credits.
How fast is it really?
fal's own model page shows an inference time of about 1.5 seconds for a 5-second 768p clip on MiniMax H3 Max Turbo, and roughly 2.7 seconds on the standard endpoint. The timer in Live Mode reports wall-clock time from the moment you press generate, which includes queueing and delivery, so it usually reads a second or two higher.
What is the highest resolution?
MiniMax H3 Max tops out at 768P — 1344×768 at 24fps in 16:9. If you need 2K, use the standard MiniMax H3 model, which renders far more slowly but at a higher resolution.
Can I use the clips commercially?
Yes. Renders you pay for are yours to use, including commercially, subject to the terms of service. Free trial clips carry the same rights.
What is the difference between MiniMax H3 Max and MiniMax H3 Max Turbo?
They are two endpoints of the same model family. Turbo renders a 5-second 768p clip in roughly 1.5 seconds and costs $0.04 per second upstream; the standard endpoint takes about 2.7 seconds and costs $0.08 per second. Turbo is the default here; the standard endpoint is the "higher quality" toggle in Live Mode.
Does it do image to video?
Yes. Both endpoints accept a first frame, and optionally an end frame, and the output follows the aspect ratio of the input image. Image to video runs on credits rather than the free trial.
What about MiniMax H3 Max Director?
Director is a separate fal endpoint. Instead of returning a finished clip, it streams video in real time and lets you send new prompts mid-session, keeping characters and scenes continuous across roughly two minutes of context. It bills per second of session with a 60-second minimum, so it is a different product shape from the per-clip generation here. This page runs the text-to-video and image-to-video endpoints; for Director sessions, go to fal directly.
Where to go next
Speed and price figures on this page come from fal's MiniMax H3 Max model pages, last checked 2026-09-03. The 75% launch promotion ends September 7, 2026; our prices are set from the regular rates, so they do not change when it does.