H3 Max

MiniMax's flagship with always-on native audio, up to 12 mixed reference files, and character tagging that keeps the same person in every shot.

Max Duration: 15 seconds max
Resolution: 480p, 768p
Aspect Ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16
Audio: Native audio

15 sec max durationUp to 12 reference filesAlways on audio6 aspect ratios

Create with H3 Max View Pricing

H3 Max aerial drone shot sweeping over a misty mountain valley at golden hour

What makes H3 Max unique

Post-trained prompt adherence

shot structure, staging, and timing follow the prompt unusually literally

Reference-to-video with mixed media

up to 12 files combined, any mix of images, video clips, and audio

Character tagging

mention Image 1 / Video 1 / Audio 1 in the prompt and each tag binds to that exact file

First-to-last keyframe control

set both endpoints for clips that cut together cleanly

Native audio, always on

dialogue, ambience, and effects are generated with every clip, no silent mode

Six aspect ratios

21:9 cinematic through 9:16 vertical, plus adaptive in reference mode

Where it falls short

–Tops out at 768p — no 1080p or 2K tier, so it is not the final-delivery pick for large screens
–Audio cannot be turned off — every clip renders with sound whether you want it or not
–Reference mode is billed at its own higher rate, plus a per-image surcharge, so it is the wrong mode for throwaway drafts
–In reference mode the aspect ratio can go adaptive, so lock it explicitly when the frame matters

Best for

Character-consistent short-form content — the same face across a whole seriesStylized transforms driven by a reference image or clipLip-sync and performance content, where the always-on audio is the pointDialogue scenes that need picture and sound composed togetherCheap, fast iteration at launch pricing before committing to a bigger model

Prompt tips

1.
Refer to reference assets by position — Image 1, Video 1, Audio 1 — and say what each one drives: 'Image 1 is the protagonist, keep her appearance consistent.'
2.
Direct the audio explicitly. It is always on, so an unmentioned soundtrack is a random one — name the dialogue, ambience, and effects you want.
3.
One reference per job: a face image for identity, a clip for the camera move, an audio file for the mood reads more clearly than twelve near-duplicates.
4.
Use a first and last frame when clips have to cut together; use reference mode when the subject has to stay consistent — they are different tools for different problems.
5.
Draft in text mode at 480p, then re-run the keeper in reference mode — reference generations cost 4x the base 768p rate plus a per-image surcharge.

How H3 Max works

Pick how you want to steer it. Text alone gets you a 5 to 15 second clip with audio at 480p or 768p. Add a first and last frame when you need exact endpoints. Or switch to reference mode and attach up to 12 files — images, video clips, and audio in any mix — then refer to them in the prompt as Image 1, Video 1, or Audio 1 to tell H3 Max exactly which file drives which part of the scene. Audio is always generated with the picture, so describe the sound you want the same way you describe the shot.

How to use on Flashloop

Step 01

Select H3 Max

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

Step 02

Write your prompt

Describe what you want to see in your video, be as detailed as you like.

Step 03

Generate & download

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

Frequently asked questions

At launch pricing (75% off until September 7): 250 credits per second at 480p and 400 at 768p, so a 5-second 768p clip is 2,000 credits. Reference mode is 1,600 credits per second plus 800 credits per reference image — a 5-second reference clip with one image is 8,800 credits.

Attach up to 12 files — any mix of images, video clips, and audio — and refer to them in the prompt by position: Image 1, Image 2, Video 1, Audio 1. Each mention binds to that exact file, so you can say 'Image 1 is the protagonist' and keep the same character across every generation that reuses that image.

No. Both are MiniMax models, but H3 Max is the newer flagship with different weights, served through a different provider, at different rates. H3 Max adds reference-to-video with mixed media and tops out at 768p; Hailuo 3 renders up to native 2K but takes no reference files.

Yes, always. Every clip is generated with native audio — dialogue, ambience, and effects are composed with the picture, and there is no silent mode. Write the soundtrack into the prompt the way you write the shot.

Any whole-second duration 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 adapt to your references.