Hailuo AI

Create short videos from a brief and reference media

Hailuo AI is MiniMax's creative video service. Its web editor brings a written brief and reference media into a workflow for generating short clips. It suits social creators, filmmakers developing a shot, and marketing teams exploring a product scene before assembling a longer production.

The current editor includes MiniMax H3 alongside creation tools and saved assets. MiniMax's H3 announcement describes a model that combines text, images, video and audio rather than treating every input as a separate task. That can help express a scene through several examples, but it does not guarantee that every reference detail will survive generation.

Choose how a shot should begin

The H3 product guide documents text-to-video, first-frame, first-and-last-frame and Omni Reference modes. A text brief is useful for a new concept; a first image gives the shot a visual starting point; two endpoint images define a transition to attempt. Omni Reference adds media cues for a character, camera movement, composition or sound.

The guide allows up to twelve reference files in total, with individual ceilings of nine images, three videos and three audio clips. An audio reference must accompany an image or video. Select only the references that explain the intended shot, and state what each should contribute instead of expecting the system to infer an entire campaign.

Sound, editing and output settings

H3 supports native stereo sound and instruction-based changes to people, objects, lighting and dialogue. Its documented clips are five to fifteen seconds, with several landscape, square and portrait ratios and output up to 2K. Check the settings offered by the selected model: the guide describes different rollout stages for resolution choices.

Targeted editing is useful when a shot is close to the brief but needs a changed object or atmosphere. It still needs review. MiniMax acknowledges remaining limitations in multimodal understanding and visual detail; an instruction to preserve a face, logo or background is an intention rather than proof that it remained exact.

An example product-teaser workflow

Suppose a small studio wants a short teaser for a desk lamp. Prepare a clear photograph of the real lamp and write a brief describing a slow camera move, a neutral desk and the light switching on. Start with one image and one intended action so the source of an unwanted change is easier to identify.

Choose the delivery aspect ratio before generating. Review the lamp's shape, the switch action, reflections and sound together. If the scene is useful but the background distracts, request that specific revision rather than replacing the entire brief. Keep the approved clip and settings with the project, then assemble titles and other shots in a video editor. This is a suggested workflow, not a claim that the lamp's identity will be reproduced perfectly.

Account access and generation cost

The web workspace provides sign-in and a credit cost beside generation controls. Usage depends on the selected mode and plan. Consult the live account offer before budgeting: promotional trials are conditional, and a temporary allowance should not be treated as an ongoing free plan. MiniMax API access is a separate route for developers, with its own platform terms and billing to review.

What should teams check before publishing?

Use reference media and voices with appropriate permission. Inspect faces, text, product geometry and sound across the complete clip, and confirm the usage terms for the account. Hailuo produces short generated shots; a finished advertisement still needs editorial assembly, factual checks and approval from the people responsible for the brand.