---
title: "MiniMax H3 AI Video Generator"
canonical: https://withicarus.com/models/h3
site: Icarus
description: "Use MiniMax H3, the video model by MiniMax, on Icarus. Sharp 2K clips from a photo or prompt."
---

# MiniMax H3 AI Video Generator

Use MiniMax H3, the video model by MiniMax, on Icarus. Sharp 2K clips from a photo or prompt.

Video model by MiniMax. A 2K clip from a prompt, a photo, or up to five references. Use it when the shot needs 2K and no multi-shot.

Try it: https://withicarus.com/dashboard/video?model=minimax%2Fh3

## What it takes

- Works from: Words alone, a start frame, a start and end frame, your photos, a clip or a sound file
- Longest description: 32,000 characters
- Reference photos: 9
- Reference clips: 3, 1.8 to 15 seconds each
- Sound files: 3, to 15 seconds each

## What it makes

- Length: 5 to 15 seconds
- Quality: 2K
- Shapes: Auto, Adaptive, 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16
- Sound: No

## Price

- 2K: $0.143 a second

Every price: https://withicarus.com/pricing#costs-heading

## Questions

### What is MiniMax H3?

MiniMax H3 is a video model made by MiniMax. A 2K clip from a prompt, a photo, or up to five references. Use it when the shot needs 2K and no multi-shot. You can use it on Icarus.

### How long can a MiniMax H3 video be?

5 to 15 seconds.

### Does MiniMax H3 make sound?

No. The clip has no sound.

### What quality does MiniMax H3 make?

2K.

### What does MiniMax H3 cost on Icarus?

$0.143 a second at 2K. You see the price before you press Generate.

### Can I use MiniMax H3 from Claude or ChatGPT?

Yes. Connect Icarus to Claude or ChatGPT once, then ask for MiniMax H3 by name in plain words. The video comes back in the chat.

## More video models

- [Seedance 2.0](https://withicarus.com/models/seedance-2.0): Video model by ByteDance. Clips from a photo or prompt.
- [Seedance 2.0 Fast](https://withicarus.com/models/seedance-2.0-fast): Video model by ByteDance. Quicker and lower priced, 480p or 720p.
- [Seedance 2.0 Mini](https://withicarus.com/models/seedance-2.0-mini): Video model by ByteDance. Lowest price per second, 480p or 720p.
- [Seedance 2.5](https://withicarus.com/models/seedance-2.5): Video model by ByteDance. Sharper motion, with sound.
- [FLUX 3](https://withicarus.com/models/flux-3): Video model by Black Forest Labs. Clips with sound, from a photo or prompt.
- [Kling 3 Motion Control](https://withicarus.com/models/kling-v3-motion-control): Video model by Kuaishou. Copy motion onto your character.
- [Kling 3.0](https://withicarus.com/models/kling-v3): Video model by Kuaishou. Clips with sound and multi-shot.
- [Kling 3.0 Omni Edit](https://withicarus.com/models/kling-o3-edit): Video model by Kuaishou. Change a clip by describing it.
- [Kling 3.0 Turbo](https://withicarus.com/models/kling-v3-turbo): Video model by Kuaishou. Quick, cheap Kling drafts, with sound.
- [LTX-2.5 Fast](https://withicarus.com/models/ltx-2.5-fast): Video model by Lightricks. Fast clips with sound, 720p to 4K.
- [LTX-2.5 Pro](https://withicarus.com/models/ltx-2.5-pro): Video model by Lightricks. Sharper clips with sound, 720p or 1080p.
- [Grok Imagine Video 1.5](https://withicarus.com/models/grok-imagine-video-1.5): Video model by xAI. Clips with sound, from a prompt, a frame, or your photos.
- [Muse Motion](https://withicarus.com/models/muse-motion): Video model by Icarus. Move your subjects the way a clip moves.
- [Muse Swap](https://withicarus.com/models/muse-swap): Video model by Icarus. Swap an object in a clip for one of yours.
- [MiniMax H3 Turbo](https://withicarus.com/models/h3-turbo): Video model by MiniMax. Quick clips from a prompt or one photo.
- [Kling Omni](https://withicarus.com/models/kling-omni): Video model by Kuaishou. A new clip from one you have, plus your photos.
- [Wan 3.0 Prime](https://withicarus.com/models/wan-3.0-prime): Video model by Qwen. Long clips with sound, from a prompt or your photos.
- [Gemini Omni 1.1 Flash](https://withicarus.com/models/gemini-omni-1.1-flash): Video model by Google. Best for UGC videos, with sound.
- [Kling 4.0](https://withicarus.com/models/kling-v4): Video model by Kuaishou. Long clips to 30 seconds, with sound.
- [Kling 4.0 Edit](https://withicarus.com/models/kling-v4-edit): Video model by Kuaishou. Change a longer clip by describing it.


---

Site: Icarus (https://withicarus.com)
Canonical: https://withicarus.com/models/h3
Contact: tanmay@hooked.design
