Why choose it
- Animate, first/last, and references
- 768p, 2K, or 4K
- Native audio
Original model creator
MiniMax
MiniMax H3
A high-detail native-audio model with first/last frames, references, and up to 4K output.
Community testSourced output
Community test · Results vary by prompt and settings
Inspect the sample frame.
Direct answer
When should you use MiniMax H3?
MiniMax H3 is a high-detail option for shots that need more visual resolution than a quick social draft. Animate a starting frame, direct both endpoint frames, or use up to nine image references at 768p, 2K, or 4K.
Best for
High-detail hero shots, atmospheric sequences, and polished brand footage.
Technical readout
The supported shot envelope.
- Duration
- 5–15 seconds
- Resolution
- 768p, 2K, or 4K
- Aspect ratio
- Landscape, portrait, and square formats
- Audio
- Native audio
What to consider
- Minimum duration is five seconds
- Higher resolutions use more credits than 768p
Prompt direction
How to prompt MiniMax H3
Use production language—lens feel, depth, material detail, light direction, and pace—to take advantage of the higher-detail output.
Start a MiniMax H3 promptQuestions, answered
MiniMax H3 FAQ
What is MiniMax H3 best for?
MiniMax H3 is best for detailed hero shots and atmospheric footage where 2K or 4K output can add value.
Can MiniMax H3 use first and last frames?
Yes. Video Lab can animate one source image or guide the shot with both a first and last frame at 768p, 2K, or 4K.
Can MiniMax H3 use reference images?
Yes. Video Lab supports up to nine MiniMax H3 reference images for subject and style guidance.
How long can MiniMax H3 clips be?
The supported range is 5 to 15 seconds.
Change the lens
Compare MiniMax H3 with another model.
Official sample
GoogleOriginal model creatorVeo 3.1
A premium text-and-image video model for cinematic shots, guided frames, references, optional audio, and up to 4K output.See model
Official sample
ByteDanceOriginal model creatorSeedance 2.5
A next-generation text-and-image video model with native audio, 30-second shots, first/last frames, and deep reference control.See model
Independent testByteDanceOriginal model creatorSeedance 2.0
A flexible text-and-image model with first/last-frame direction, up to nine references, broad formats, and optional audio.See modelMiniMax lens selected
Direct your first MiniMax H3 shot.
Write the prompt, choose supported controls, and see the exact credit quote before you generate.
Open MiniMax H3 in Video Lab