What Is Hailuo H3? MiniMax Video Model Guide
Lukas Vogel
Applied Research Editor

TLDRHailuo H3 launched on July 31, 2026 with 15-second 2K video, native audio, multimodal reference control, and video editing. Platform access and pricing details vary.
Inside Hailuo H3: 15-Second 2K Clips at Seedance 480p Pricing
Hailuo H3 is a launched AI video generation model from MiniMax that generates clips up to 15 seconds long at 2560×1440 resolution with native audio, multimodal reference control, and consistent characters across scenes. MiniMax teased the release on July 30, 2026 under the tag #MiniMaxH3, and H3 launched on July 31, 2026. Access is now available through several first-party and third-party surfaces, including MiniMax Design and MiniMax Hub, while platform-specific pricing and limits vary.
Key Takeaways
- Hailuo H3 is MiniMax's launched video model, teased on July 30, 2026 and launched on July 31, 2026 under the #MiniMaxH3 tag.
- Confirmed capabilities include 15-second maximum clip length, 2560×1440 (2K) resolution, native audio, multimodal reference control, and character consistency across scenes.
- Community pricing reports include approximately $1 per 15-second 2K clip on a basic subscription, while another usage report lists a $10.50 Standard plan sale price, 204 credits for a 10-second clip, and ¥320 for a 10-second clip.
- H3 is MiniMax's first major video model release in roughly eight months, according to a tester citing the gap since the prior Hailuo model.
- Video-to-video editing works from simple natural-language instructions in hands-on demonstrations.
- H3 is available through MiniMax Design, MiniMax Hub, Topview, PoYo API, Fal, and additional third-party services, but a complete official pricing and API specification remains unclear.
What Is Hailuo H3?
Hailuo H3 is the next iteration of MiniMax's Hailuo video generation model line, positioned as a text-to-video and image-to-video model with native audio and integrated editing. It is also called MiniMax H3. Launch materials and hands-on tests describe text, image, video, and audio reference control, along with first-and-last-frame control and video editing.
H3 launched on July 31, 2026, after first surfacing publicly through a coordinated set of teasers. MiniMax's official account posted a preview clip with the caption "It's coming" on July 30, launch partner WaveSpeedAI published a spec-heavy sneak peek, and creator AshutoshShrivastava confirmed early-access status the same morning.
The model is presented under two overlapping brand tags — Hailuo H3 on the product side via @Hailuo_AI and MiniMax H3 on the corporate side. Both refer to the same model.
Hailuo H3 at a Glance
| Attribute | Value |
|---|---|
| Developer | MiniMax (product brand: Hailuo AI) |
| Type | Video generation model |
| Modality | Text-to-video, image-to-video, reference-to-video, video-to-video editing |
| Max clip length | 15 seconds |
| Max resolution | 2560×1440 (native 2K) |
| Native audio | Yes, including stereo audio in reported platform access |
| Character consistency | Yes, across scenes |
| Reference control | Text, image, video, and audio; reported support for up to 9 images, 3 videos, and 3 audio clips |
| Pricing | Reported at about $1 per 15s 2K clip on a basic subscription; other platform figures vary |
| API availability | Yes, through PoYo API and third-party API providers |
| Availability | Launched July 31, 2026; available through MiniMax Design, MiniMax Hub, Topview, PoYo API, Fal, and other providers |
| License | Not yet confirmed |
| Model card / paper | Not yet published |
How Hailuo H3 Works and What Makes It Different
Three feature anchors define H3 in the material available after launch: Omni Reference, Native Audio, and Scene-Persistent Characters. Omni Reference allows multimodal conditioning across text, images, video, and audio. Launch-related platform posts also describe reference inputs, including up to nine images, three videos, and three audio clips in one reported configuration.
Native Audio means that audio and video are generated together rather than requiring a separate dubbing step. H3 is described as producing native audiovisual output, with stereo audio reported on platform access. Creator tests also mention dialogue, sound effects, music synchronization, and lip-sync alignment, although these observations are not standardized benchmarks.
Scene-Persistent Characters is the third pillar. H3 demonstrations describe consistent characters across different scenes, and creator comparisons frequently praise identity retention and visual continuity. That makes the model useful for connected short-form sequences where the same people or designs must carry through multiple shots.
The 2560×1440 ceiling matters because it places H3 in the native 2K class of video generators. Combined with the 15-second length, a single H3 job can produce a full short-form scene rather than a stitched sequence of 4-6 second fragments.
"A 15 second 2k resolution clip costs $1 on the global site, going off the basic sub price. Seedance 1080p, same length would cost $4 on Dreamina with the basic sub." — Instagram creator alvia.art, community pricing report
What You Can Do With Hailuo H3
Post-launch demonstrations cluster into three use cases:
-
Cinematic short-form video from text prompts. WaveSpeedAI's preview highlights "cinematic visuals" as a core aesthetic target, and AshutoshShrivastava's first test generated four clips from a single prompt that he then stitched together in post.
-
Video-to-video editing from natural-language instructions. Alex Patrascu posted a demo captioned "Hailuo H3 can edit your footage really well! Simple instructions, crazy good", showing existing footage modified via short text prompts.
-
Anime and stylized short-form content with native sound. Creator demonstrations include anime-styled horror shorts generated with H3 and native audio, while other tests use detailed prompts, reference images, and multiple media inputs to control characters, scenes, and sound.
The common thread across these demonstrations is that H3 is positioned less as a raw text-to-video engine and more as an end-to-end short-form video tool: prompt → visual → audio → light editing.
The reference workflow is particularly important for commercial and serialized content. Product-commercial tests report smooth motion and consistent product design from a single image, while other creators have tested character sheets, multiple reference images, and connected scenes. H3 also supports first-and-last-frame control in reported launch materials, giving creators another way to define how a shot begins and ends.
How Hailuo H3 Compares
Comparison points here are drawn from community tests rather than standardized benchmarks. Treat as directional.
| Dimension | Hailuo H3 | Seedance 2.0 (reference) |
|---|---|---|
| Max resolution | 2560×1440 (native 2K) | 1080p in one comparable pricing report |
| Max clip length | 15 seconds | 15 seconds in comparable creator reports |
| Reported cost, 15s clip | About $1 at 2K in one basic-subscription report; another report says roughly 30% of Seedance 2.0 pricing | About $4 at 1080p in the cited basic-subscription comparison |
| Large sweeping camera moves | Weaker in one creator comparison | Stronger in that comparison |
| Native audio | Yes | Not established by the comparisons summarized here |
| Reference retention and lip sync | Favored by some testers | Favored by some testers for skin and facial detail or cinematic expression |
The takeaway from community reactions is mixed rather than settled. One same-prompt comparison favored H3 for sharper light, cleaner color, and more motion detail. Another preferred Seedance 2.0 as more cinematic and expressive and less blurry or morphed. A later comparison gave H3 the edge on reference retention and lip sync but Seedance 2.5 the edge on skin texture and fine facial changes.
H3's clearest reported advantage is price, particularly for 2K output, while its performance on large sweeping camera movements and high-action cinematography remains contested. For context on where Seedance itself has moved recently, see our Seedance 2.5 release deep dive.
Availability: How to Access Hailuo H3
Hailuo H3 launched on July 31, 2026, and is no longer limited to the small group of pre-launch creative partners. Access has been reported through MiniMax Design and MiniMax Hub, as well as launch and inference platforms including Topview, PoYo API, and Fal.
API-style access is also available through third-party services that list H3, including GoEnhance and Pixazo. EvoLink describes H3 as live with day-one access and offers generation of 4-15 second clips in native 2K with stereo sound. These services may apply their own pricing, credit systems, account requirements, and geographic restrictions.
The current access surfaces are:
- MiniMax Design and MiniMax Hub — creator-facing surfaces where H3 has been launched and demonstrated.
- Topview and PoYo API — launch-era platform access, including multimodal generation and stereo audio reports.
- Fal and other third-party providers — additional access points for generation or API workflows, including services listing GoEnhance, Pixazo, and EvoLink support.
Teams building comparable short-form video pipelines can also work with the previous-generation Hailuo 2.3 via API.
What We Don't Know Yet
Hailuo H3 has launched, but several technical and commercial details remain unpublished or incompletely verified:
- Official pricing tiers for subscriptions, API usage, credits, and usage limits. Community reports include about $1 for a 15-second 2K clip, a $10.50 Standard plan sale price, 204 credits for a 10-second clip, and ¥320 for a 10-second clip, but these figures come from different access contexts.
- Parameter count, architecture family, and whether it shares a backbone with prior Hailuo models.
- An official MiniMax API endpoint, model ID string, and rate limits. Third-party API access exists, but provider-specific listings do not establish MiniMax's own API documentation.
- License terms for generated content and any commercial-use restrictions.
- Formal benchmark scores on standard video generation evaluations.
- Whether weights will be released. One single-source report describes H3 as open-weights, but that claim is not corroborated across the available material.
- Universal availability and geographic coverage. H3 is accessible through multiple products and providers, but the available reports do not establish identical access for every region or account type.
- The full scope of editing controls, including how all video-editing operations are exposed across platforms.
The launch resolves H3's release status, but not every question about its implementation, pricing, or long-term distribution.
Frequently Asked Questions
What is Hailuo H3?
Hailuo H3 is a launched AI video generation model from MiniMax, the company behind the Hailuo video product line. It generates clips up to 15 seconds long at 2560×1440 resolution with native audio, multimodal reference control, and consistent characters across scenes. MiniMax teased the model on July 30, 2026, and H3 launched on July 31, 2026.
Who makes Hailuo H3?
Hailuo H3 is made by MiniMax. The model ships under the Hailuo AI product brand and is also called MiniMax H3, including under the #MiniMaxH3 tag.
How much does Hailuo H3 cost?
Hailuo H3 pricing varies by platform, and a complete official tier breakdown has not been established. Users have reported about $1 for a 15-second 2K clip on a basic subscription, while another usage report listed a $10.50 Standard plan sale price, 204 credits for a 10-second clip, and ¥320 for a 10-second clip.
Is Hailuo H3 available to use?
Yes. Hailuo H3 launched on July 31, 2026, with access reported on Topview, PoYo API, and Fal, as well as MiniMax Design and MiniMax Hub. Third-party access is also available through services listing H3 APIs or generation, including GoEnhance, Pixazo, and EvoLink, although geographic availability and provider-specific limits may vary.
What resolution and length does Hailuo H3 support?
Hailuo H3 supports generations up to 15 seconds at 2560×1440 resolution, also described as native 2K. It produces native audio, supports multimodal references, and maintains character consistency across different scenes.
How does Hailuo H3 compare to Seedance 2.0?
Community comparisons are mixed. Some testers favored Hailuo H3 for sharper light, cleaner color, motion detail, reference retention, or lip sync, while others preferred Seedance 2.0 for cinematic quality and expressive motion. H3 is also repeatedly described as less expensive, with one report putting its cost at roughly 30% of Seedance 2.0 pricing, but there is no standardized cost or quality benchmark.
Does Hailuo H3 support video editing?
Yes. Hailuo H3 demonstrations show it accepting existing footage plus simple text instructions and editing it in place. One tester described the results as "crazy good" from short natural-language prompts, though full editing controls have not been documented.
What to Watch Next
Three signals will clarify H3's position as the launched product matures. First, an official MiniMax pricing breakdown — that is when the different per-clip, subscription, and credit reports can be reconciled. Second, official MiniMax API documentation showing the model ID, endpoint, and rate limits, since third-party API access already exists but does not establish the vendor's full API specification. Third, side-by-side benchmark posts against Seedance 2.5 and Kling 3.0 from creators who have access to all three; those will reveal where H3 actually places on motion and prompt adherence rather than on cost alone.
Building similar short-form video with native audio? On kie.ai you can try MiniMax H3, Seedance 2.5, and Hailuo 2.3.
About Lukas Vogel
Lukas reads the papers and model cards so you do not have to, focusing on reproducible claims.
View all posts by Lukas Vogel