The cheapest MiniMax H3 is here,$0.013/s
Flux 3 Max Video Generator
Text to video and image to video on one post-trained model, with a 10-second clip back in about 12 seconds
None

None

None

Long Story Video Skill

Turn your idea into 10s–10min video—AI writes scripts, prompts, and generates footage automatically.

Ads Video Skill

Ads Video Skill

Generate professional ads and sales videos—AI auto-generates scripts, prompts, and footage.

3D Science Explainer Video Skill

Convert scientific concepts into stunning 3D explain animations

AI Video Prompt Generator
10s

Feedback

Flux 3 Max

Flux 3 Max by fal Research: #1-tier image-to-video quality and a 10-second clip in about 12 seconds. Try it at 720p, 75% off until November 7.

All Tools

Discover our comprehensive AI-powered animation toolkit

Why Use Flux 3 Max

Flux 3 Max is fal Research's post-trained model on FLUX 3, built with Black Forest Labs for top-tier quality at near real-time speed.

  • Post-trained on the FLUX 3 base
    fal Research kept training Black Forest Labs' FLUX 3 further, and is still refining it.
  • Tied for #1 in image-to-video
    It scores 1,139 Elo in Artificial Analysis's preliminary arena, level with the leaders and ahead of Wan 3.0.
  • 10-second clips in about 12 seconds
    Ten seconds of video arrives in roughly twelve, near real time, at quality other models need minutes to reach.

How to Use Flux 3 Max

Pick a mode, run a prompt in the playground, or connect Flux 3 Max through the fal API.

Flux 3 Max Features

Close-up faces, motion that keeps its shape, faithful prompts, and a place at the fast end of the quality-latency chart.

Faces that hold up in close-up

A low angle on the station keeper: skin, sweat-stained felt and pale sky stay believable where most models blur.

Motion that stays true to life

Speeding horses, spinning wheels and drifting dust keep their shape across the whole clip, with no warping mid-shot.

Prompt fidelity on unfilmed shots

Pitch a booster landing on a frontier ranch at dawn and the whole idea lands, scale, dust and dawn light included.

Fastest corner of the chart

On a rating-versus-speed chart it lands in the fastest corner, next to H3 Max, with no rival scoring higher while moving quicker.

Three layers of work at once

Tuning, serving and systems work advanced side by side, because each gain opened room in the next, so quality and speed rose together.

Swap models with one model ID

Both fal Research models live on one API, so moving a shot from Flux 3 Max to H3 Max is a one-line ID change.

FAQ

Flux 3 Max — FAQ

What Flux 3 Max is, how post-training works, how fast it runs against Seedance 2.5 and H3 Max, and what it costs.

1

What exactly is Flux 3 Max and who made it?

A fal Research video model built with Black Forest Labs on FLUX 3: Black Forest Labs made the base, fal added training, serving and systems work.

2

What does post-training mean?

It starts from a finished base and keeps training it on new data; the base knows motion, light and framing, while post-training sharpens prompts and polish.

3

How fast is it next to Seedance 2.5 and H3 Max?

Ten seconds of video in 12.45 seconds, versus 277.80 for Seedance 2.5 at the same 1,139 Elo. H3 Max is quicker still, at 7.55 seconds.

4

How is Flux 3 Max priced?

Per second of output, no minimums, no subscription. 75% off until November 7 brings it to $0.0425 per second at 720p; list is $0.17, both modes alike.

5

How does it compare with plain FLUX 3?

The post-trained version gains 50 Elo over FLUX 3 (1,089 to 1,139) and cuts a ten-second render from 126.44 seconds to 12.45, about ten times quicker.

6

Can output be used in commercial projects?

Yes. Work generated through the fal.ai API can go into commercial projects; the terms of service spell out the licensing details.

Run Flux 3 Max in the Playground

Open a pre-loaded prompt and generate, or send a first request with the fal SDK. 75% off until November 7, billed per second with no subscription.