5s · 768P · from text
Live quote
audio incl.
AI video · MiniMax
Direct the scene. Shape the sound.
MiniMax H3 creates 5–15-second videos from text, opening or ending frames, or mixed references, at resolutions from 480P to 4K. Direct identity, motion and native stereo sound together, or add your soundtrack in text and image modes.
Three creation workflows
Move between text-to-video, image-to-video and reference-to-video inside one H3 engine.
5–15 seconds at up to 4K
Choose any whole-second duration and render at 480P, 768P, 2K or 4K at 24 FPS.
Multimodal reference direction
Combine image, video and audio references to guide identity, performance, motion and sound.
Native stereo audio
Use a soundtrack in text or image mode, or audio references in reference mode. Describe dialogue, ambience and synchronized sound cues in the prompt.
Choose workflow, duration, resolution and references, then review the exact price before rendering.
Live quote
audio incl.
Live quote
Most popularaudio incl.
$2.88
audio incl.
$1.95
audio incl.
The first five reference images are included; each additional reference image adds its displayed surcharge.
Use text-to-video and choose Auto or a fixed aspect ratio for a complete shot from one prompt.
Use a start image, an end image or both. Frame guidance and mixed-reference inputs are separate workflows.
Use reference mode to assign identity, wardrobe, motion and voice cues across the scene.
Build the prompt in three passes: define the performance, direct the shot sequence, then lock continuity and sound.
Source: MiniMax H3 guide
Describe the subject, action, camera and sound. Add an optional soundtrack to guide the audio.
Use a start image, an end image or both, with an optional soundtrack. Keep frame inputs separate from mixed references.
Use up to 9 images, 3 videos and 3 audio clips within 12 unique references. Audio-only references work too; video and audio each have a 15-second combined limit.
Use a soundtrack in text or image mode, or audio references in reference mode. Describe dialogue, ambience and synchronized sound cues in the prompt.
Establish identity, expression, body language and the visible change in the scene.
Character: [Appearance, wardrobe, expression and voice] Setting: [Location, time, weather and visual anchors] Action: [One clear physical action and reaction] Sound: [Dialogue, ambience and synchronized effects]
Performance defines one readable character action and emotional change.
Best practices, common fixes, and important limitations to help you get the strongest results with H3.
Compare prices, formats, and capabilities before choosing your model.
Use Kling 3.0 Omni Pro for reference-guided Kling video, storyboard images, @Image prompt anchors, optional start/end frames, video-to-video, and native audio.
Compare H3 vs Kling 3.0 Omni ProGenerate cinematic Veo 3.1 videos with text prompts, start-image animation, multi-reference guidance, optional last-frame control, and extend workflows in one unified MaxVideoAI model page.
Formats, settings and limits to check before creating.
Built-in safeguards and best practices for responsible creation with H3.
Use a text prompt, a start image with an optional end image, or a reference set containing images, videos and audio. Audio references must accompany at least one image or video reference.
Choose any whole-second duration from 5 through 15 seconds and render at 768P, 2K or 4K at 24 FPS.
Reference mode accepts up to 9 images, 3 videos and 3 audio clips, with no more than 12 unique references in one request.
Yes. H3 produces native stereo audio with the video; there is no separate audio switch.
Yes. MiniMax H3 is available in MaxVideoAI, and the exact price is shown before every render.