How Hailuo AI Turns One Prompt Into Nine Kinds of Video

Chapter 1

One Screen, Twelve Empty Reference Slots

A pet owner opens hailuoai.video expecting to hunt through menus before finding anything usable. Instead there is one large text box, sitting almost alone on the page, waiting for a sentence.

Above the box sits a small placeholder line – "A child flying a kite in the park, golden sunlight, camera tilting up." – which is not an example someone has to copy, it is simply showing the shape of a prompt: a subject, a setting, a mood, a camera movement. Next to the box are two labelled chips: one names the model doing the work, MiniMax H3, the other reads Omni Reference. Beside them sits a counter: Refs (0/12).

That counter is the first thing worth understanding before typing anything, because it changes what the box is actually capable of. Twelve is not a decoration – it is a limit on how many of your own images this one screen will accept before you have even written a word.

What you can actually do here

Hailuo's generation screen is identical wherever you land on the site – what changes is which door you walked through to get there. Nine separate pages exist purely to point a specific kind of maker toward the same tool.

Nine doors, one generator

Chapter 2

What Filling Those Slots Actually Buys You

Ask any model to generate the same face, the same pet or the same product twice from a written description alone, and it will usually hand back two different-looking results. That is the ordinary failure of text-only generation: nothing anchors it to a specific subject, so each attempt drifts.

The twelve reference slots exist to fix exactly that. Upload a handful of real photos of the subject – a dog, a couple, a baby, a product – and the model has something concrete to hold onto rather than guessing from adjectives. Hailuo calls this pairing Omni Reference, and it sits directly beside the prompt box rather than buried in a settings menu, which suggests it is meant to be used on the very first attempt, not discovered later.

The model behind it, MiniMax H3, is described on the site as doing native multimodal generation and precise multimodal editing – in plain terms, it is built to take more than plain text as its input and to let a person adjust a generation rather than only regenerate it from scratch. Nothing on these pages demonstrates that editing step directly, so it is worth treating as a stated capability rather than a proven one until you have tried it yourself.

Once a subject is anchored through its references, the next question is what the finished clip will actually look like – and that is decided by four smaller settings sitting just below the prompt box.

Chapter 3

The Four Dials Below the Prompt

Four small controls sit in a row under the text box: 2K, 5s, 21:9, and a count of 1. Read together, they are the resolution, the length, the shape, and the number of clips a single generation will produce.

2K sets the output sharpness. 5s sets how long the clip runs – short enough that a single generation is a moment, not a scene, which matters for planning: an ad, a meme or a love-video clip built this way is closer to a highlight than a story told start to finish. 21:9 sets the frame as a wide, cinema-style rectangle rather than the vertical shape most phone video is shot in, which is worth noticing before building something meant for a phone feed. The final number controls how many versions of the same prompt are produced in one pass.

Next to the Create button sits one more number: 60. Nothing on these pages spells out what it measures, but its position – directly beside the button that starts a generation, not beside the reference slots or the resolution setting – reads as a cost tied to running the generation itself, most likely in whatever credit system sits behind the free allowance covered in the next chapter. Worth confirming for yourself before assuming what it buys.

📰 Recommended for You

How Canva Turns an Indian Occasion Into a Printed Invitation

Read More →

How Animaker Turns One Prompt Into a Finished Video

Read More →

None of these four settings appeared with any alternative values on the pages captured here – no toggle to a longer clip, no second resolution option, no portrait frame. Whether those exist elsewhere in the interface once logged in is not something these pages show, so it is better treated as unconfirmed than assumed absent.

Chapter 4

Nine Doors Into the Same Room

Type "hailuoai.video" into a browser and the generation screen above is what appears. But the site also carries at least nine separate pages, each built around one specific kind of clip someone might want to make, each opening onto the exact same prompt box, reference slots and four dials.

An ad generator page speaks to a small business owner building marketing footage. A pet video page speaks to someone who wants their dog or cat animated rather than still. A meme generator page speaks to someone making something quick and shareable rather than polished. A photo-transformation page speaks to someone who has one photo and no written idea at all – it leads with turning a still image into motion, prompt optional. A style-transfer page speaks to anime and art audiences wanting a rendered look rather than a realistic one. A love-video page speaks to couples working from relationship photos, and a baby-story page speaks to new parents doing the same with a child's photos. An ASMR page speaks to creators after something slow and textural rather than fast-cut. And a cinematic page names something none of the others do – a "Director Mode" – though it appears only in that one page's title, with nothing else on the site explaining what it changes.

None of this changes what the tool does. It changes what a visitor sees first, and which examples and framing greet them. A reader arriving through the pet page and a reader arriving through the ad page land on the same reference slots and the same dials – the only real decision left is which door matches the clip already in mind, which is exactly what the map in this book is for.

Chapter 5

The Free Credits, and What Isn't Shown Yet

Every page captured here carries the same banner: download MiniMax Design and receive 3,000 free credits to start creating. It is the only figure on the entire site tied to cost, and it is worth treating as a launch offer rather than a permanent one – tied, by its own wording, to a separate app download rather than to simply using the website.

Once those credits run out, nothing on these pages says what comes next. There is no visible pricing page, no plan comparison, no note about what a subscription would include. That gap sits in the site's own documentation, not in what this book chose to leave out – a reader deciding whether this tool fits their budget should check the pricing page directly before committing time to a project built here.

The same is true of a tutorials section named on every page – "MiniMax H3 Tutorials: Open-Weight Projects, Hands-on Tutorials, and Creative Workflows" – which is pointed to consistently but whose own content sits behind a link this book has not walked through. It is real, it is named clearly, and it is worth opening directly rather than guessing at from a homepage banner.

What is clear, from the screen itself, is simpler and more useful for a first attempt: twelve reference slots to anchor a subject, four dials that fix resolution, length, shape and count, and nine doors that lead to the same room. Pick the door that matches the clip already in mind, load a few real photos into the reference slots if the subject needs to look like something specific, and let the dials decide the rest before spending a credit finding out the hard way.

Contact / More useful information from RamthaMedia

Official source links:
Hailuo AI