Updated

MiniMax

MiniMax H3 Max

fal's post-trained version of MiniMax H3: 5 to 15 second clips with native audio and up to 12 reference files.

Create with MiniMax H3 MaxView pricing

What is MiniMax H3 Max?

MiniMax H3 Max is a video model that fal.ai built on top of MiniMax's open-weight H3, the model MiniMax sells as Hailuo 3.0.

fal post-trained the weights for closer prompt following and faster generation. On Flashloop it makes 5 to 15 second clips at 480p or 768p, with native audio in every render.

Reference mode is where it earns its place. You can attach up to 12 files, mixing images, short video clips and audio, and point at each one in the prompt as Image 1, Video 1 or Audio 1. The model ties each mention to its file, which is how a character keeps the same face from one generation to the next. fal reports it renders a 5 second 768p clip in about three seconds, roughly 35 times the throughput of the official MiniMax H3 endpoint.

Use it for short series, lip-sync and dialogue scenes, where sound and a consistent cast matter more than pixel count. It stops at 768p on Flashloop, so move approved shots to a model like Veo 3.1 or Kling 3.0 when you need 1080p. Audio is always generated, which helps with dialogue and gets in the way if you plan to score the clip yourself.

Provider
MiniMax
Length
15 seconds
Resolution
480p, 768p
Aspect ratios
21:9, 16:9, 4:3, 1:1, 3:4, 9:16
Audio
Native audio

What you can do with MiniMax H3 Max

  • Close prompt following

    fal post-trained it so shot structure, staging and timing land where you write them

  • Mixed reference files

    up to 12 images, video clips and audio files in one generation

  • Reference tagging

    call a file Image 1, Video 1 or Audio 1 in the prompt and the model ties the mention to that file

  • First and last frame

    set both ends of a clip so consecutive shots cut together

  • Native audio on every clip

    dialogue, room tone and effects come out of the same pass as the picture

  • Six aspect ratios

    21:9, 16:9, 4:3, 1:1, 3:4 and 9:16

Prompts to try with MiniMax H3 Max

“Medium two-shot in a small ramen shop at night. Image 1 is the chef, Image 2 is the customer. The customer says "Same as last time?" and the chef nods and slides a bowl across the counter. Steam rising, low jazz from a radio, rain on the window. 10 seconds, 16:9.”

Try this prompt

Where MiniMax H3 Max fits best

  • Episodic shorts with a recurring cast

    Tag the same face image as Image 1 in every prompt and the character carries across episodes without being described again.

  • Talking-head and lip-sync clips

    Speech is generated in the same pass as the picture, so mouth movement and dialogue line up without a dubbing step.

  • Restyling footage from a reference

    Attach a clip as Video 1 for the motion and an image as Image 1 for the look, then describe how the two combine.

  • Shots that have to cut together

    Set a first and last frame so each clip starts and ends exactly where the edit needs it.

  • Quick concept passes

    Draft at 480p in text mode to test staging and timing, then rerun the take you like with references.

How MiniMax H3 Max compares

MiniMax H3 Max compared with other video models on Flashloop
ModelLengthTop resolutionAspect ratiosNative audio
MiniMax H3 Max15 seconds768p21:9, 16:9, 4:3 +3Yes
Veo 3.18 seconds1080p16:9Yes
Kling 3.015 secondsPro (1080p)16:9, 9:16, 1:1No
Kling 3.0 Turbo15 seconds1080p16:9, 9:16, 1:1No
Kling 4.0Coming soon30 secondsTo be confirmedTo be confirmedYes

Good to know before you generate

  • On Flashloop it stops at 768p, so use another model for 1080p finals
  • Audio is always generated and there is no silent mode
  • Reference mode costs more per second than text mode, so test the idea in text mode first
  • With references attached the aspect ratio can follow the inputs, so set it yourself when framing matters

Start creating with MiniMax H3 Max

Create with MiniMax H3 Max

  1. Select MiniMax H3 Max

    Open Flashloop and choose MiniMax H3 Max from the model selector in the video creator.

  2. Enter your prompt

    Describe what you want to see in your video.

  3. Generate and download

    Hit generate and your video will be ready in seconds. Download or share directly.

Questions about MiniMax H3 Max

How much does H3 Max cost on Flashloop?

H3 Max is billed per second of video. The rate goes up with resolution, and reference mode costs more than text or frame mode. The creator shows the exact credit cost before you generate.

How does reference tagging work in H3 Max?

Attach up to 12 files in any mix of images, video clips and audio, then refer to them by position in the prompt: Image 1, Image 2, Video 1, Audio 1. Each mention is tied to that file, so writing 'Image 1 is the protagonist' keeps the same character in every generation that reuses the image.

Is H3 Max the same model as MiniMax Hailuo 3?

No, but they are related. MiniMax released H3, sold as Hailuo 3.0, with open weights in July 2026. fal.ai post-trained those weights into H3 Max for closer prompt following and faster generation. On Flashloop, H3 Max adds mixed-media references and stops at 768p, while Hailuo 3 renders up to 2K.

Does H3 Max generate audio?

Yes, on every clip. Dialogue, ambience and effects are generated with the picture and there is no silent mode, so write the soundtrack into the prompt the same way you write the shot.

What durations, resolutions and aspect ratios does H3 Max support?

Any whole-second length from 5 to 15 seconds, at 480p or 768p, in 21:9, 16:9, 4:3, 1:1, 3:4 or 9:16. In reference mode the ratio can also follow your reference files.

Who makes H3 Max?

fal.ai. MiniMax built and open-sourced the H3 base model, and fal Research post-trained it into H3 Max together with the MiniMax team, releasing it on August 27, 2026. Flashloop gives you access to it alongside other video models.

How is H3 Max different from H3 Max Turbo?

H3 Max Turbo is fal's faster, distilled version of H3 Max. It keeps native audio and first and last frame control but drops reference mode, so it cannot use tagged images, clips or audio. Pick H3 Max when a character has to stay consistent and Turbo for quick drafts.

More video models on Flashloop

Browse all models