# AI models on MiniMax H3

Canonical page: https://minimax-h3.io/models

> Compare MiniMax H3 (Hailuo 3.0), the platform's core AI video model, with supporting Kling, Veo, Seedance, Wan, Grok Imagine, Nano Banana, GPT Image, Seedream, and Flux options

## Overview

MiniMax H3 (Hailuo 3.0) is the core video model on this platform. It supports text-to-video, image-to-video, synchronized dialogue and sound, flexible aspect ratios, and clips up to 15 seconds. Other video and image models remain available as supporting choices for projects that need a different output profile, control set, or cost.

Credit costs vary by model, resolution, duration, and optional features. The values below match the current generation controls and may change when provider pricing changes.

## Core Video Model

### MiniMax H3 (Hailuo 3.0)

- **Best for**: The platform's primary text-to-video and image-to-video workflow
- **Audio**: Synchronized dialogue, sound effects, and ambience
- **Resolutions**: 768P, 2K
- **Aspect Ratios**: 1:1, 4:3, 3:4, 9:16, 16:9, 21:9
- **Duration**: Any whole second from 4s through 15s
- **Reference Images**: First/last-frame image-to-video; the provider reference mode accepts up to 9 images

| Resolution | Cost |
|---|---:|
| 768P | 22 credits/s |
| 2K | 36 credits/s |

The first five reference images are included. Each additional image costs 11
credits. Input video is billed at the selected resolution's per-second rate;
input audio is free. The current generator UI accepts text and first/last-frame
images, and does not yet submit video or audio references.

## Supporting Video Models

### Kling 3.0

- **Strengths**: Native audio, first/last-frame image-to-video, and output up to 4K
- **Resolutions**: 720p, 1080p, 2160p
- **Aspect Ratios**: 1:1, 9:16, 16:9
- **Duration**: 5s, 10s, 15s

| Resolution | Without audio | With audio |
|---|---:|---:|
| 720p | 40 credits/s | 50 credits/s |
| 1080p | 50 credits/s | 70 credits/s |
| 2160p | 190 credits/s | 190 credits/s |

### Veo 3.1

- **Strengths**: Google's quality tier with native audio for text-to-video and image-to-video
- **Resolutions**: 720p, 1080p
- **Aspect Ratios**: 16:9, 9:16
- **Duration**: 4s, 6s, 8s

| Resolution | Cost per video |
|---|---:|
| 720p | 700 credits |
| 1080p | 720 credits |

### Veo 3.1 Fast

- **Strengths**: Faster, lower-cost Veo tier with native audio
- **Resolutions**: 720p, 1080p
- **Aspect Ratios**: 16:9, 9:16
- **Duration**: 4s, 6s, 8s

| Resolution | Cost per video |
|---|---:|
| 720p | 160 credits |
| 1080p | 180 credits |

### Seedance 2.5

- **Strengths**: ByteDance's newest video model, with synchronized audio, first/last-frame image-to-video, and clips up to 30 seconds
- **Resolutions**: 480p, 720p, 1080p
- **Aspect Ratios**: 1:1, 4:3, 3:4, 9:16, 16:9, 21:9
- **Duration**: 4s, 5s, 8s, 10s, 15s, 20s, 30s

| Resolution | Cost |
|---|---:|
| 480p | 70 credits/s |
| 720p | 160 credits/s |
| 1080p | 400 credits/s |

### Seedance 2.0

- **Strengths**: Text-to-video and image-to-video across a wide range of aspect ratios
- **Resolutions**: 480p, 720p, 1080p
- **Aspect Ratios**: 1:1, 4:3, 3:4, 9:16, 16:9, 21:9
- **Duration**: 4s, 5s, 8s, 10s, 15s

| Resolution | Cost |
|---|---:|
| 480p | 50 credits/s |
| 720p | 100 credits/s |
| 1080p | 230 credits/s |

### Seedance 2.0 Fast

- **Strengths**: Faster generation for prompt testing and iteration
- **Resolutions**: 480p, 720p
- **Aspect Ratios**: 1:1, 4:3, 3:4, 9:16, 16:9, 21:9
- **Duration**: 4s, 5s, 8s, 10s, 15s

| Resolution | Cost |
|---|---:|
| 480p | 40 credits/s |
| 720p | 80 credits/s |

### Wan 3.0

- **Strengths**: Alibaba's flagship video model for long, low-cost clips, with an optional audio track and first/last-frame image-to-video
- **Resolutions**: 480p, 720p, 1080p
- **Aspect Ratios**: 1:1, 4:3, 3:4, 9:16, 16:9
- **Duration**: 5s, 10s, 15s, 20s, 30s

| Resolution | Cost |
|---|---:|
| 480p | 20 credits/s |
| 720p | 40 credits/s |
| 1080p | 80 credits/s |

Wan 3.0 costs the same with or without audio.

### Grok Imagine

- **Strengths**: xAI's fast video model with sound in every clip, from text or one image; the lowest cost per second on the platform
- **Resolutions**: 480p, 720p, 1080p
- **Aspect Ratios**: 1:1, 9:16, 16:9 (with an image, the clip keeps the image's proportions)
- **Duration**: 4s, 5s, 6s, 8s, 10s, 12s, 15s

| Resolution | Cost |
|---|---:|
| 480p | 7 credits/s |
| 720p | 12 credits/s |
| 1080p | 22 credits/s |

Grok Imagine always generates sound; the price is the same for text and image input.

## Supporting Image Models

These models create standalone images, campaign assets, references, and source frames that can feed into a MiniMax H3 image-to-video workflow.

### Nano Banana 2.1

- **Strengths**: Google's successor to Nano Banana 2, with readable in-image text and steady edits
- **Resolutions**: 1K, 2K, 4K
- **Aspect Ratios**: 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9
- **Reference Images**: Up to 10 in the MiniMax H3 workspace

| Resolution | Cost per image |
|---|---:|
| 1K | 20 credits |
| 2K | 20 credits |
| 4K | 30 credits |

### Nano Banana 2

- **Strengths**: Fast, reasoning-guided image generation for rapid iteration
- **Resolutions**: 1K, 2K, 4K
- **Aspect Ratios**: 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9
- **Reference Images**: Up to 5 in the MiniMax H3 workspace

| Resolution | Cost per image |
|---|---:|
| 1K | 30 credits |
| 2K | 40 credits |
| 4K | 60 credits |

### Nano Banana Pro

- **Strengths**: Google's highest-quality image tier for detailed and artistic output
- **Resolutions**: 1K, 2K, 4K
- **Aspect Ratios**: 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9
- **Reference Images**: Up to 5 in the MiniMax H3 workspace

| Resolution | Cost per image |
|---|---:|
| 1K | 60 credits |
| 2K | 60 credits |
| 4K | 80 credits |

### GPT Image 2.5

- **Strengths**: OpenAI's newest image model, with sharper detail and consistent people and products across edits
- **Resolutions**: 1K, 2K, 4K
- **Aspect Ratios**: 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9
- **Reference Images**: Up to 8 in the MiniMax H3 workspace

| Resolution | Cost per image |
|---|---:|
| 1K | 20 credits |
| 2K | 40 credits |
| 4K | 60 credits |

### GPT Image 2

- **Strengths**: Prompt-accurate image generation and strong text rendering
- **Resolutions**: 1K, 2K, 4K
- **Aspect Ratios**: 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9
- **Reference Images**: Up to 5 in the MiniMax H3 workspace

| Resolution | Cost per image |
|---|---:|
| 1K | 20 credits |
| 2K | 40 credits |
| 4K | 60 credits |

### Seedream 5.0 Pro

- **Strengths**: ByteDance's flagship image model for dense text, data-rich layouts, posters, and infographics
- **Resolutions**: 1K, 2K
- **Aspect Ratios**: 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9
- **Reference Images**: Up to 8 in the MiniMax H3 workspace

| Resolution | Cost per image |
|---|---:|
| 1K | 30 credits |
| 2K | 50 credits |

### Grok Imagine Image

- **Strengths**: xAI's image model for photoreal scenes and quick edits
- **Resolutions**: 1K (single tier)
- **Aspect Ratios**: 1:1, 2:3, 3:2, 9:16, 16:9
- **Reference Images**: Up to 5 in the MiniMax H3 workspace

| Resolution | Cost per image |
|---|---:|
| 1K | 20 credits |

### Flux 2 Pro

- **Strengths**: Photorealistic detail, prompt following, and multi-reference consistency
- **Resolutions**: 1K, 2K
- **Reference Images**: Up to 8

| Resolution | Cost per image |
|---|---:|
| 1K | 20 credits |
| 2K | 30 credits |

### Flux 2 Flex

- **Strengths**: Higher-control composition, typography, and fine detail
- **Resolutions**: 1K, 2K
- **Reference Images**: Up to 8

| Resolution | Cost per image |
|---|---:|
| 1K | 50 credits |
| 2K | 80 credits |

## How to Choose

- **Start with MiniMax H3** for the platform's main text-to-video or image-to-video workflow, synchronized audio, and clips up to 15 seconds.
- **Choose Kling 3.0** when first/last-frame control or 4K output is the deciding factor.
- **Choose Veo 3.1** for Google's native-audio video workflow, or Veo 3.1 Fast for a lower-cost alternative.
- **Choose Seedance 2.5** for ByteDance's newest model and clips up to 30 seconds, or Seedance 2.0 when you need a lower-cost supporting model with broad aspect-ratio coverage.
- **Choose Wan 3.0** for long clips up to 30 seconds at a low cost per second.
- **Choose Grok Imagine** for quick clips with sound at the lowest cost per second.
- **Choose Nano Banana 2.1, 2, or Pro** for fast image iteration or higher-fidelity Google image output.
- **Choose GPT Image 2.5** for OpenAI's sharpest, most consistent image output, or GPT Image 2 for prompt-accurate images and text rendering.
- **Choose Seedream 5.0 Pro** for text-heavy posters, infographics, and layouts.
- **Choose Flux 2 Pro or Flex** for photorealistic multi-reference work or tighter composition control.

## Questions?

If you have questions about MiniMax H3, supporting models, or pricing, contact us at [support@minimax-h3.io](mailto:support@minimax-h3.io) or through the [contact page](https://minimax-h3.io/contact).