AI Video Generator
Models
Sora Series

OpenAI's latest video generation model with enhanced quality

10 credits
10s/15s
Portrait/Landscape
Veo Series

Google's latest model with exceptional quality output

150 credits
Up to 8s
16:9/9:16
Seedance Series

30-second multimodal video generation with precise reference control and native audio

0 credits
Up to 30s
7 Aspect Ratios
10s
Public Visibility

When this option is enabled, the output video may be selected by AI Generate and published to the Explore.

Notification
This is a notification message
Sample Videos

MiniMax H3 · Lovart ME

MiniMax H3 AI Video Generator

4–15s · 768P / 2K

Create 4–15 second videos from text, first and last frames, or a reference package. Choose 768P or 2K output. Billing includes generated seconds and input-video seconds; the first five reference images are free, with 4 credits for each additional image.

4–15s
Duration
768P / 2K
9
Reference images
Use a clean source image and describe the protected details, requested changes and final composition.
768P / 2K
Resolution
768P: 8/s · 2K: 13/s
3
Text and image inputs
Create in Lovart ME, then review the result in your history.

01 / MODEL

What you can make · MiniMax H3

Create in Lovart ME, then review the result in your history.

Text-to-video with a chosen ratio

Describe the scene and choose one of six fixed aspect ratios. Keep the subject action, environment and camera direction readable within 4–15 seconds.

First and last frame animation

Upload an opening image, or an opening and ending pair. These images guide the endpoints of the shot. The frame workflow follows source framing and does not expose a separate aspect-ratio field.

A reference package for appearance and motion

Reference mode accepts up to nine images and three video clips. Up to three audio files can accompany an image or video; audio cannot be the only reference.

Estimate the complete job

At 768P use 8 credits/s; at 2K use 13 credits/s. Multiply by generated plus input-video seconds, then add 4 credits for each reference image beyond five. Input audio is free.

02 / INPUTS

Choose the input that carries useful direction

Create 4–15 second videos from text, first and last frames, or a reference package. Choose 768P or 2K output. Billing includes generated seconds and input-video seconds; the first five reference images are free, with 4 credits for each additional image.

Text and image inputs

Use text for a new scene. Upload references only when they provide useful visual direction.

Reference images

9 · Use a clean source image and describe the protected details, requested changes and final composition.

Prompt

State the action or edit, composition and lighting. Give each reference one role.

03 / OUTPUT

Set the output before generating

Choose a supported resolution, ratio and duration where available. Review the full cost before submitting.

Resolution

768P / 2K

Credits

768P: 8/s · 2K: 13/s Credits

Aspect ratios

21:9, 16:9, 4:3, 1:1, 3:4, 9:16

04 / PROMPTS

MiniMax H3 · Prompts for practical production tasks

Use a clean source image and describe the protected details, requested changes and final composition.

Portrait action

Adapt the subject, setting and protected details to your own brief.

PromptA musician sits beside a window and opens a wooden instrument case. Six-second continuous shot, soft afternoon light, slow camera push-in. Keep the hands and instrument geometry clear.

Working note Use a clean source image and describe the protected details, requested changes and final composition.

Frame-controlled reveal

Adapt the subject, setting and protected details to your own brief.

PromptOver six seconds, move from the supplied opening frame to the ending frame. Preserve the room, subject identity and light direction. Make the transition a single continuous camera movement.

Working note Use a clean source image and describe the protected details, requested changes and final composition.

Reference motion

Adapt the subject, setting and protected details to your own brief.

PromptUse the reference image for the cyclist and clothing. Use the reference video only for camera tracking. Generate a six-second shot of the cyclist riding along a quiet coastal road. Preserve the reference appearance.

Working note Use a clean source image and describe the protected details, requested changes and final composition.

05 / WORKFLOW

How to use MiniMax H3

Create in Lovart ME, then review the result in your history.

  1. 01

    Choose text, frames or references

    Use text for a new scene. Upload references only when they provide useful visual direction.

  2. 02

    Describe the subject and what must stay fixed

    State the action or edit, composition and lighting. Give each reference one role.

  3. 03

    Set output and review the displayed price

    Choose a supported resolution, ratio and duration where available. Review the full cost before submitting.

  4. 04

    Generate, inspect and revise

    Inspect details and continuity, then change one or two instructions per attempt. Your job remains available in history.

06 / FAQ

MiniMax H3 · Questions before you generate

Create in Lovart ME, then review the result in your history.

What can I create with MiniMax H3?

Create 4–15 second videos from text, first and last frames, or a reference package. Choose 768P or 2K output. Billing includes generated seconds and input-video seconds; the first five reference images are free, with 4 credits for each additional image.

Which output settings are available?

Resolution: 768P / 2K. Duration: 4–15s. Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16.

How is the price calculated?

768P costs 8 credits/s and 2K costs 13 credits/s, applied to generated plus input-video duration. The first five reference images are free. Each additional image costs 4 credits; audio input is free. Up to three reference videos with a combined duration of 15 seconds. Audio requires an accompanying image or video.

What should reference uploads contain?

The first five reference images are free. Each additional image costs 4 credits; audio input is free. Up to three reference videos with a combined duration of 15 seconds. Audio requires an accompanying image or video.

How long does generation take?

Generation has no fixed completion time. Queue load, output settings and upstream availability affect the wait.