Private AI

Nano GPT logo
NanoGPT
Nano GPT logo
NanoGPT
Nano GPT logo
NanoGPT
Search
 
 
 
 
All conversations

Create

Chat
Studio

Explore

Models
Pricing
AI Apps
Blog

Library

Gallery
Workspace
Usage & billing
API
Suggestions and bugs
Use cases

|

Terms

|

Privacy

|

Refunds
Usage & billing
Teams
Referrals
Presets
Settings
FAQPress & mediaBrand assetsBenchmarksEarn crypto
  1. Models
  2. Video

MiniMax H3

Try Model

Generate from text, images, or references with optional custom LoRAs. Editing and extension require no LoRA weights.

Added Jul 31, 2026

Details

Starting Price
From $0.120 per video
Model type
both

Final price depends on the selected settings and is shown before generation.

Settings

Generation controls available for this model.

Output Format

up to 1,080p
Auto
Square
Portrait
Landscape

Default Duration

5

13 duration options

Aspect Ratio

Select

Default

auto

Options (8)

auto, 21:9, 16:9, 4:3 +4 more

Auto uses 16:9 for LoRA generation.

Duration (seconds)

Select

Default

5

Options (13)

3, 4, 5, 6 +9 more

Generate Edit Audio

Toggle

Default

Yes

LoRA / Edit / Extend Resolution

Select

Default

480p

Options (4)

480p, 540p, 768p, 1080p

LoRA generation supports all listed resolutions; edit/extend support 480p and 768p. Generation without LoRAs remains 2K.

LoRA 1 strength

Number

Default

1

LoRA 1 URL

Text

Default

N/A

HTTPS URL of a compatible LoRA weights file.

LoRA 2 strength

Number

Default

1

LoRA 2 URL

Text

Default

N/A

HTTPS URL of a compatible LoRA weights file.

LoRA 3 strength

Number

Default

1

Showing 9 of 15 settings.

Benchmarks

Human preference benchmarks sourced from LMArena.

Text to Video

#11 / 50

Arena Score

1456.2

Votes

29,442

Confidence Interval

1448.2 - 1464.2

Image to Video

#3 / 50

Arena Score

1492.6

Votes

122,064

Confidence Interval

1487.1 - 1498.1

Video Edit

#4 / 11

Arena Score

1382.8

Votes

8,088

Confidence Interval

1373.9 - 1391.7

Published 2026-11-08 · Matched as MiniMax H3

LMArena Dataset

Examples

Loading examples…

Related video models

Compare MiniMax H3 with similar models from the same provider or model family.

MiniMax H3 Highspeed

minimax/h3-highspeed

MiniMax H3 Highspeed creates 4–15 second videos at 768p with native audio. Generate from text, animate a starting image with an optional end frame, or combine up to nine image, three video, and three audio references. Reference clips must be 2–15 seconds each, totaling at most 15 seconds per media type. Audio references require an image or video. Frame guidance and reference inputs cannot be combined.

MiniMax H3 Lite

minimax/h3-lite

MiniMax H3 Lite creates 4–15 second videos at 480p or 768p with native audio. Generate from text, animate a starting image with an optional end frame, or combine up to nine image, three video, and three audio references. Reference clips must be 2–15 seconds each, totaling at most 15 seconds per media type. Audio references require an image or video. Frame guidance and reference inputs cannot be combined.

MiniMax H3 Max Turbo

minimax/h3-max-turbo

MiniMax H3 Max Turbo generates 480p, 768p, or 1080p video from text, first/last frames, or image/video/audio references with synchronized audio. Text and frame modes support 5–15 seconds; references support 0.92–15 seconds and may run about 0.7 seconds longer. Use up to 9 images, 3 videos, and 3 audio clips, with 12 files total and 2–15 seconds per clip, at most 15 seconds per media type. Reference generation reserves $5 from your account balance and refunds the unused amount after billed output and reference usage are confirmed.

MiniMax H3 Spicy Image-to-Video

wavespeed-ai/minimax-h3/image-to-video-spicy

Open-weights MiniMax H3 image-to-video generation with expressive unrestricted motion, native stereo audio, optional last-frame control, 3–15 second clips, and 480p or 768p output.

MiniMax H3 Max

minimax/h3-max

Generate from text, first/last frames, or image/video/audio references.

MiniMax Hailuo 2.3 Pro

minimax-hailuo-23-pro

MiniMax Hailuo 2.3 Pro delivers cinematic 1080p 5-second clips with advanced physics, prompt fidelity, and character consistency. Supports both pure text prompts and image-conditioned motion.