Hailuo AI Video Generator

Hailuo is an AI video model family for short generated clips. Reviral currently lists Hailuo 2.3, Hailuo 2.3 Fast, and Hailuo 02. Choose a text-capable entry or add the eligible image required by Fast, select a supported duration and resolution, review the live credit charge, then generate a clip for a short-form scene or ad edit.

The current controls are compact and specific. Hailuo 2.3 and Fast list six or ten seconds at 768P, while 1080P is available only for six seconds. Standard accepts up to one image; Fast requires one. The entries expose prompt optimization and a watermark toggle, but no generate-audio control is listed, so plan separate sound work when the final edit needs it.

Six seconds is the common option across 768P and 1080P for Hailuo 2.3 and Fast. Ten seconds is listed at 768P, not 1080P. That makes the decision concrete: choose six seconds when higher listed resolution is required, or ten seconds when the action needs more time and 768P fits the delivery plan. Do not select a duration and resolution independently without checking the valid combination.

Standard Hailuo 2.3 accepts zero or one reference image. Fast sets a minimum of one and a maximum of one, so it is an image-led entry. Hailuo 02 accepts up to two, although its 512P option allows zero or one. One image can anchor a subject, product, palette, or opening composition, but the prompt still needs to specify the action, camera, and ending.

The current Hailuo registry entries do not include a generate-audio extra or an audio-resolution list. That is a meaningful absence. Use the clip as a visual element and plan dialogue, narration, captions, ambience, music, and effects in the surrounding production workflow. Avoid describing the model as a native-audio generator unless a later visible entry and registry update prove that control.

Use a compact prompt with five jobs: identify the subject, locate it in a setting, describe one action in order, direct the camera, and state the final frame. For an image-led Fast request, say what should remain recognizable from the image and what must change. A specific ending pose or product position gives you a clean review target.

Match complexity to duration. Six seconds can establish one subject and complete a simple action. Ten seconds can give movement more setup and resolution, but it does not turn the clip into a multi-scene story. Split location changes, multiple claims, or unrelated actions into separate generations. That makes failures cheaper to diagnose and successful shots easier to reuse.

The registry caps these prompts at 2,000 characters and enables a prompt optimizer by default. The optimizer does not remove the need for a coherent brief. Avoid contradictory camera directions, invisible backstory, and unsupported performance claims. Decide whether a watermark belongs in the request, review that toggle, and inspect the completed frame rather than assuming the setting was applied.