How to Use Hailuo 02: MiniMax's Physics-First Video Model on VdoBloom

Hailuo 02 is MiniMax’s physics-and-camera model on VdoBloom: it generates 6 or 10 second clips from either a text prompt or a starting image, at 768P or 1080P, costing 39 to 45 credits per generation, and it carries the catalog’s permissive HOT content label rather than the strict filtering used on the Hailuo 2.3 tiers. That combination — dual input modes, real physics simulation, and looser filtering — makes it the Hailuo model most people should reach for first.

Disclosure: VdoBloom is our platform. Competitor details were checked against their own site and we name the competitor as the better pick wherever that is true.

What Hailuo 02 is actually best at

Most AI video models fall apart on physics. Liquids hang in the air, hair passes through shoulders, thrown objects drift instead of arcing. Hailuo 02 is built around next-generation physics simulation and camera control, and that shows up specifically in shots involving weight, momentum, and contact: something falling and landing, water breaking over a rock, fabric caught in wind, a body pivoting and following through.

The second strength is deliberate camera work. Hailuo 02 responds well to explicit camera direction in the prompt — a slow push-in, an orbit around the subject, a low-angle rise — rather than inventing random drift. If your shot list has camera moves written into it, this model will honour them more reliably than models tuned mainly for pretty static scenes.

It also does both jobs in one model. Give it a prompt alone and it generates a scene from scratch; upload a still and it animates from that frame. On VdoBloom that means the same model serves the text-to-video and image-to-video workflows, which is unusual in the Hailuo family — the 2.3 tiers are image-only.

Real specs and exact credit costs

The option matrix is deliberately small: two durations, two resolutions, and one combination that does not exist. Hailuo 02 does not expose an aspect-ratio control — framing comes from your prompt on text-to-video, or from the source image on image-to-video. At the Lite plan rate of 300 credits for $15, one credit is $0.05, so the dollar figures below are exact.

Duration	Resolution	Credits	Cost at $0.05/credit	What it is for
6 seconds	768P	39	$1.95	The default. Prompt tests, motion checks, most social clips.
6 seconds	1080P	45	$2.25	Delivery quality at the shorter length. Only 6 more credits than 768P.
10 seconds	768P	45	$2.25	Longer beat, same price as 1080P at 6s. Choose length over sharpness.
10 seconds	1080P	Not available	—	The model does not support 1080P at 10 seconds.

That table hides a genuinely useful pricing quirk. The jump from 768P to 1080P at 6 seconds costs 6 credits — 30 cents. The jump from 6 to 10 seconds at 768P costs the same 6 credits. So once you are past the draft stage you are effectively choosing between resolution and length at an identical price, and there is no option that buys you both. Decide which one your clip actually needs before you spend.

How to prompt Hailuo 02 well

Because the model’s edge is physics and camera, wasted prompt words are the ones describing static beauty. Spend them on movement instead. A structure that works:

• Subject and state — who or what, and what condition they start in. “A surfer crouched low on a board, knees loaded.”
• The physical event — the thing with weight and consequence. “The wave collapses over the nose of the board, spray tearing sideways.”
• Camera instruction — named and singular. “A slow tracking move alongside, staying level with the rider.”
• Light and grade — one line, no more. “Hard late-afternoon backlight, high contrast.”

Three habits that reliably improve results: give the camera exactly one instruction per clip, because stacking a push-in with an orbit and a tilt in six seconds produces mush; describe the follow-through of a motion, not just the motion, since that is what the physics engine renders convincingly; and avoid asking for dialogue or lip movement, which is not this model’s strength.

If you are animating a photograph of a real person, you must have that person’s consent before you generate — that applies to every image-to-video model on the platform, not just this one.

A worked example

Say you need a 10-second hero clip of a coffee pour for a cafe ad. Start at 6 seconds and 768P for 39 credits ($1.95) with a prompt like: “Dark espresso pours from a portafilter into a clear glass, crema building and swirling as the stream hits; a slow push-in centred on the glass; warm window light from the left, shallow depth of field.” Look at one thing only in the result — does the liquid behave like liquid? If yes, re-run the identical prompt at 10 seconds and 768P for 45 credits ($2.25) to get the longer pour.

Total spend: 84 credits, $4.20, for a tested draft and a finished clip. The mistake to avoid is opening at 1080P: you pay 45 credits to discover the pour physics were wrong, then pay 45 again for the fix.

When to pick a different model

Hailuo 02 is not the right answer for every shot. If your source is a finished piece of art direction — key art, a product photograph, a rendered character — and you simply want it animated at the highest rendering quality, Hailuo 2.3 Image-to-Video Pro is the image-only tier built for that, at 25 credits for 6 seconds at 768P. If you are running many motion tests on one image, Hailuo 2.3 Standard starts at 15 credits ($0.75) per clip, well under half the Hailuo 02 entry price.

The trade-off runs the other way on content filtering. Both 2.3 tiers are labelled STRICT and screen your uploaded image as well as your prompt. Hailuo 02 carries the HOT label, meaning more permissive filtering — useful for stylised combat, horror atmospheres, swimwear, dance and fitness concepts that stricter models refuse. Explicit and illegal content stays blocked platform-wide regardless of model; see the flexible-filter overview for where that line sits.

If you want MiniMax’s Director mode with named camera moves and Subject Reference for character consistency, MiniMax’s own Hailuo AI product does that better than prompt-level camera direction here — our VdoBloom vs Hailuo AI comparison says so plainly. What VdoBloom adds is that Hailuo 02 sits in one credit pool alongside 105+ other models, and one-time credit packs from $2.49 never expire, while Hailuo’s own credits reset monthly. Full plan details are on the pricing page.

Frequently asked questions

How much does one Hailuo 02 clip cost?

39 credits for 6 seconds at 768P, and 45 credits for either 6 seconds at 1080P or 10 seconds at 768P. At the Lite rate of 300 credits for $15, that is $1.95 and $2.25 respectively. New accounts get 10 free credits, which is not quite enough for one Hailuo 02 clip — test the cheaper models first if you are on the free tier.

Can Hailuo 02 make a 10-second 1080P video?

No. The model supports 1080P at 6 seconds and 768P at both 6 and 10 seconds, but there is no 10-second 1080P option. If you need a longer clip at full resolution, generate 10 seconds at 768P and upscale afterwards, or plan the shot as two 6-second 1080P clips.

Does Hailuo 02 need a starting image?

No — it works from text alone or from an image. That makes it different from the Hailuo 2.3 Pro and Standard tiers, which are image-to-video only and will not run without a source frame.

Why is Hailuo 02 labelled HOT and 2.3 labelled STRICT?

The labels describe the provider’s content filtering, not the content itself. Hailuo 02 runs more permissive filters, so edgier lawful concepts get through more often. The 2.3 tiers apply strict screening to both the source image and the prompt. Explicit and illegal material is blocked on every model.

Can I set an aspect ratio on Hailuo 02?

There is no aspect-ratio control for this model. On text-to-video, describe the framing you want in the prompt; on image-to-video, the source image’s shape drives the output, so crop your still to the ratio you need before uploading.