MiniMax teased its next video model on July 30, 2026 with a two-word post — "It's coming" — and the internet immediately started searching for a tool that does not exist yet. Hailuo 3.0 (officially tagged Hailuo H3 on the product side, MiniMax H3 on the corporate side — both names refer to the same model) is real, previewed, and impressive on paper. It is also not publicly available. This guide covers what is actually confirmed, how access works right now, and what to use while you wait.
What Hailuo 3.0 actually is
Hailuo H3 is the third major numbered generation in MiniMax's video model line, following Hailuo 02 and Hailuo 2.3. The Hailuo 3.0 preview surfaced on July 30, 2026 through a coordinated teaser: MiniMax's official account posted a preview clip, launch partner WaveSpeedAI published a spec sneak peek, and a handful of early-access creators confirmed they were already testing it.
Here is everything that has been stated publicly so far — and nothing more:
| Attribute | Confirmed preview spec |
|---|---|
| Developer | MiniMax (product brand: Hailuo AI) |
| Modes | Text-to-video, image-to-video, video-to-video editing |
| Max clip length | 15 seconds |
| Max resolution | 2560×1440 (2K) |
| Native audio | Yes — generated with the video, not dubbed after |
| Character consistency | Yes, across scenes in one session |
| Pricing | Reportedly ~$1 per 15s 2K clip on a basic subscription — unconfirmed |
| API | Not yet announced |
| Availability | Early access only, small pool of creative partners |
Three features anchor the preview material. Native audio means the soundtrack is generated jointly with the footage rather than layered on afterward — one of the three headline upgrades early testers list over Hailuo 2.3. Scene-persistent characters means a subject's identity carries from one shot to the next within a generation session, which has been a long-standing weak point of every video model. Omni Reference is a multi-modal reference-conditioning system that early testers say they are still exploring; treat detailed claims about it as premature.
One useful context point: testers note this is MiniMax's first major video release in roughly eight months. The company skipped the incremental-update cadence its competitors kept, which is why expectations around Hailuo H3 are unusually high.
How to use Hailuo 3.0 today — the honest answer
You mostly can't. As of early August 2026 there is no public app toggle, no API, no pricing page, and no model card. Access is limited to invited creative partners. Anything presenting itself as "Hailuo 3.0 free online" right now is one of three things:
- A squatter site. Exact-match domains for the model name went up before the model did. They run older models (or nothing) behind the new name.
- A rebadged Hailuo 2.3. Several platforms quietly serve the current model under next-gen branding. 2.3 is genuinely good — but it is not Hailuo H3, and it does not have native audio.
- A waitlist page. Legitimate, but a waitlist is not a tool.
When access does open, expect the official Hailuo app and the Hailuo web studio to get it first — that was the rollout order for Hailuo 2.3, with API platforms following shortly after. If you want to be early through legitimate channels, two moves actually work. Follow the official Hailuo AI account, where early-access rounds have been announced. And watch the API aggregators — MiniMax's previous models reached third-party API platforms shortly after public launch, and per-second pricing there has historically undercut the native subscription for casual volumes.
What to use while Hailuo 3.0 is early-access
The confirmed Hailuo H3 specs are a useful shopping list, because most of them exist today in other models — just not all in one place.
For 2K-class fidelity right now, Kling 3.0 is the closest currently-available match in our studio, with strong prompt adherence and optional audio. For character-led clips from a photo, Seedance 2.0 handles image-to-video with reliable subject preservation. For quick, free experiments, our video studio opens on a free engine — useful for prototyping the prompt you plan to run on Hailuo H3 later, because prompt craft transfers between models far more than people expect.
Two workflow habits worth building before Hailuo H3 arrives:
- Write shot lists, not vibes. Every Hailuo H3 preview clip that impressed came from structured direction — subject, camera move, lighting, beat timing. If your prompts already read like a proper shot list, you will be productive on Hailuo H3 from the first generation.
- Practice the extend workflow. Hailuo H3's 15-second ceiling is generous, but stories still outrun it. Chaining a clip past its last frame with a video extender is the technique that turns any per-clip cap into a soft limit, on Hailuo H3 the same as anywhere else.
Reading the Hailuo 3.0 price signal
The one economic data point in circulation — roughly $1 for a 15-second 2K clip on the basic tier, reported by an early-access partner and not confirmed by MiniMax — is worth taking seriously as a signal even if the number moves. It would price Hailuo H3's 2K output at approximately what Seedance 2.0 charges for 480p at the same length. If that holds at launch, it repositions the mid-tier of the whole video-model market, and the platforms reselling per-generation credits will have to reprice within weeks.
The counter-scenario is equally plausible: early-partner pricing is often promotional, and native-audio generation is compute-heavy. Do not budget a production pipeline around an unconfirmed number.
Hailuo 3.0 FAQ
Is Hailuo 3.0 the same thing as Hailuo H3 and MiniMax H3?
Yes. The company uses Hailuo H3 in both brandings; "Hailuo 3.0" is the community's version-number phrasing. All three names describe one model.
Is Hailuo 3.0 free to use?
No pricing has been announced, and there is no public free tier because there is no public anything. Previous Hailuo releases did include limited free daily generations on the official app, so a free lane at launch is plausible — but it is speculation until MiniMax says so.
Can I get the Hailuo 3.0 feature set somewhere else today?
Individually, mostly yes: 2K-class output, image-to-video, and audio exist across current models — see the alternatives above. The genuinely new combination is native audio plus scene-persistent characters in one generation. Nothing shipping today does both at once.
When will Hailuo 3.0 launch publicly?
MiniMax has not said. The teaser pattern — official clip, partner spec sheets, creator early access — is the same sequence the company ran before its previous launches, where public release followed within weeks rather than months. That is an observed pattern, not a promise.
Should I wait for Hailuo H3 or build with what exists?
Build now. Prompt structure, reference-image habits and the extend workflow all transfer to whatever you switch to later. The people who get the most out of a new model in week one are always the ones who arrived with a working pipeline.
We track model launches like this one closely because our studio integrates new engines shortly after their APIs stabilize. Specs above are drawn from MiniMax's July 30 teaser materials and early-partner reports; anything unconfirmed is labeled as such, and we will update this page when the model actually ships.
