MiniMax H3

Turn ideas and references into product ads and cinematic scenes with native audio.

Enter your idea to generate
MiniMax H3

Key Features of MiniMax H3

MiniMax H3 brings scene direction, visual references and sound into one video generation workflow. Here are five ways to use it. Images are AI-generated concept illustrations of these workflows.

Text to video

Build an ad or story scene from an idea

Describe who is in the shot, what happens and how the camera moves. Turn a written brief into a product teaser, an establishing shot or a short story moment without preparing source footage. For a scene like a traveler waiting for a train, specify the approaching headlights, slow camera push and rainy evening atmosphere.

Image references

Give your product or character a visual starting point

Use product photos, character portraits or location images to show H3 what the scene should draw from. A studio bottle photo can guide a product reveal in a garden; a portrait can guide the subject in a new setting. Combine references and explain which details matter, then describe the action you want to add.

Video references

Show the movement that is hard to describe

Use a short clip to communicate a dance phrase, a gesture or a camera move. Pair it with an image reference to give the subject and movement separate visual guidance. Explain what to borrow from the clip and what to change—useful when a written description alone leaves too much of the timing or staging unclear.

First and last frames

Plan where the shot begins and where it ends

Animate a single image when you already have the opening composition. Add a last frame when the destination matters too: a closed box becoming a product reveal, a pose changing, or a storyboard moving to its next beat. H3 generates the movement between your visual endpoints; describe the transition and keep the two frames visually compatible.

Native audio

Direct the scene and its sound together

Describe dialogue, music, ambient sound or effects alongside the visible action. H3 generates audio with the video, so you can explore a guitar performance, a rainy cafe or a spoken scene in one pass. In reference mode, add audio together with an image or video and explain how it should guide the scene.

How to Use MiniMax H3

01

Choose what will guide the shot

Start with text alone in reference mode, or add images, video and audio. Choose first and last frames when you want to animate an opening image or connect two compositions. Keep these two reference modes separate.

02

Describe the action, camera and sound

Write a clear scene brief: subject, setting, action, camera movement and desired sound. Explain each reference's role. Choose 4–15 seconds and 768P or 2K; select the output shape for reference mode, while frame mode follows the image.

03

Check the cost, generate and refine

Review the credit estimate before generating. For an inexpensive first test, use 4 seconds at 768P with text or images. Review the motion and sound together, refine the brief, then choose a longer duration or 2K when the direction works.

FAQ About MiniMax H3

We've answered the most frequently asked questions

Use H3 for product teasers, social ads, character scenes, animated images and short storyboard shots. You can start from a prompt, animate a first frame, connect first and last frames, or combine image, video and audio references to guide a new scene.

Use references to guide a subject, product, setting or style while allowing a new composition. Use a first frame when the video should start from your image; add a last frame to guide its ending. These modes cannot be combined in one generation. On LumiYing, a last frame needs a first frame.

Reference mode accepts up to 9 images, 3 videos and 3 audio clips. Each video or audio clip must be 2–15 seconds; videos may total at most 15 seconds, and audio may separately total at most 15 seconds. Audio references need at least one image or video. Use a prompt to explain how the files should work together.

H3 supports 768P and 2K, with whole-second durations from 4 to 15 seconds. For text and references, choose 16:9, 9:16, 4:3, 3:4, 1:1 or 21:9. First and last frame generation follows the input image's aspect ratio. Choose portrait for vertical social clips or landscape for wider scene compositions.

Yes. H3 generates native audio with the video. Describe the dialogue, ambience, effects or music you want in the prompt. Audio clips can also guide a reference-based scene when paired with visual media. Review the generated sound with the picture before using the result.

768P costs 8 credits per billable second; 2K costs 12. Billable time is output duration plus the combined input video duration, rounded to the nearest whole second. A 5-second output without video references costs 40 or 60 credits. Add a 10-second reference video and the same output costs 120 or 180 credits. LumiYing measures input video duration before charging.

Start with a 4-second, 768P shot using text or images: 32 credits. The same shot at 2K costs 48 credits. Keep the scene simple and test one action first. Video references increase billable time, so only add them when their movement or timing is needed.

This H3 page creates new videos from text, frames or references. An uploaded reference video guides a new generation; it is not an edit or extension of the original clip. Video editing and continuation are not available for H3 in the current LumiYing generator.

Choose the scene. Direct what happens.

Start from a brief, a product photo or a pair of storyboard frames. Use MiniMax H3 to explore the movement, composition and sound of your next short video.