MiniMax H3 —
AI Video Generator with Native Audio
MiniMax now lists H3 as available, with confirmed 4–15 second output, 768P and 2K options, and multimodal references. Generate on this site, or read our API guide for the workflow, supported formats, and site pricing.
Native Stereo Audio
Official API: 4–15s
Text + Image Inputs
Free to Start
MiniMax H3 Video Examples & Prompts
Use these examples as prompt-study material: note the subject, action, camera, lighting, continuity, and sound intent. A useful MiniMax H3 prompt explains what changes over time, not just the first frame. Open the prompt library to adapt proven MiniMax H3 structures for text-to-video, image-to-video, dialogue, and reference-led scenes.
MiniMax H3 Text to Video & Image to Video
MiniMax H3 Text to Video
Describe a complete shot in natural language: subject, setting, action, camera, lighting, pacing, dialogue, and ambience. MiniMax H3 turns that direction into a short cinematic sequence, while follow-up controls help refine characters, objects, backgrounds, and spoken lines.
MiniMax H3 Image to Video
Upload a still image to define composition, character, product, or the first frame. MiniMax H3 image-to-video adds camera and subject motion while using the source as a visual anchor. It is useful when a team already has approved artwork, a storyboard frame, or a product photo and wants motion without losing the original direction.
Build the Prompt Around a Shot
Strong results start with a clear hierarchy: name the main subject first, then movement, environment, framing, lighting, sound, and the intended ending. Prompts can be detailed, but concise direction is easier to revise. Start with one controlled scene before expanding into a multi-shot sequence.
Guide Identity with References
Reference media can carry details that are awkward to repeat in prose: a face, outfit, location, voice, rhythm, or camera language. Images, video, and audio can give the generated clip more context for consistency. Upload only media you are allowed to use and keep the most important identity cue clear.
Review Motion and Sound Together
Evaluate a generated clip as a connected audiovisual result. Check whether the subject stays recognizable, actions complete cleanly, lip movement matches dialogue, sound effects land at the right moment, and the final frame supports the next edit. If one element fails, revise that instruction rather than rewriting every successful part of the prompt.
MiniMax H3 Features: Native Audio, Omni-Reference & Editing
The official API confirms multimodal generation from text, images, video, and audio references. Other creative claims below remain provider-dependent unless the current official documentation confirms them.
Native Stereo Audio & Lip Sync
MiniMax H3 is described as generating sound with the picture: dialogue, ambience, music, and timed effects. Native generation can reduce the mismatch of adding audio after rendering, while lip sync and voice handling may help spoken scenes feel connected. The current V2 endpoint confirms reference-audio input, but output-audio behavior can still vary by provider, prompt, and language.
Omni-Reference
Use reference images, video, or audio to guide identity, art direction, movement, voice, and atmosphere. More files do not automatically improve the result; choose references that reinforce one clear MiniMax H3 video concept and check the limits shown for the selected mode.
Instruction-Based Editing
Instruction-based editing can target characters, objects, backgrounds, effects, and dialogue. Instead of describing the whole scene again, ask for one bounded change, such as new weather, a slower camera, a replaced prop, or a rewritten line.
Use cases
MiniMax H3 Use Cases
The model can support compact production tasks where visual references, motion direction, and sound concepts need to stay connected. Begin with short, reviewable shots and treat the official 4–15 second API range as a testing envelope rather than a delivery guarantee.
Content Creators & Social Media
Create short social concepts, talking-character tests, product teasers, and visual hooks from a written brief. Native audio can establish dialogue or atmosphere during ideation, while an image reference can hold a recognizable host, mascot, or art direction. Exported results should still be reviewed for timing, claims, captions, rights, and platform-specific requirements before publication.
Marketing Teams & E-Commerce
Turn approved product photography or campaign key art into motion studies before commissioning a full shoot. A generated video can test camera movement, pacing, sound design, and localized dialogue, helping stakeholders compare directions quickly. Keep packaging text and factual product claims under human review, because generative details can drift between frames.
Filmmakers & Studios
Use MiniMax H3 for previsualization, mood pieces, transition tests, and shot blocking. References can carry wardrobe, location, camera language, performance rhythm, and a temporary sound bed. Short clips work well for testing one dramatic beat before an editor combines approved shots.
Developers & API Integration
Prototype generator interfaces, creative review queues, prompt libraries, and media-management workflows around short audiovisual outputs. The official V2 endpoint uses model ID MiniMax-H3, while third-party provider identifiers may differ. Validate authentication, rate limits, safety behavior, file constraints, pricing, and actual output settings before treating any integration as production-ready.
What Is MiniMax H3 (Hailuo 3)?
MiniMax H3 is now officially available through MiniMax's video generation V2 API. The official endpoint supports text-to-video, first- and last-frame image-to-video, and reference-to-video with image, video, and audio inputs. MiniMax H3 is also searched as Hailuo 3 and MiniMax 3, but it is the video model—not the MiniMax M3 language model or Tencent Hy3.
MiniMax H3's official API confirms 768P and 2K output, 4–15 second integer durations, a 7,000-character prompt limit, and up to nine reference images, three reference videos and three reference audio files (12 files total). On 3 August 2026 MiniMax published the open weights (33B parameters), and the official model card adds 24 FPS and 32 kHz stereo audio. Note that the 1440p figure circulating in the community is not an official tier — the official options are 768P and 2K.
MiniMax H3: Available API and Current Specs
Status
Official API available
Resolution & frame rate
768P & 2K · 24 FPS
Clip duration
4–15 seconds (integer)
Audio
32 kHz native stereo
Omni-reference
Up to 12 total files
Prompt length
Up to 7,000 characters
MiniMax H3 vs Hailuo 2.3, Sora 2 & Veo
This is a practical planning comparison, not a benchmark. MiniMax H3 combines native audio, multimodal references, and focused editing; Hailuo 2.3 is an earlier generation in the same family, while Sora 2 and Veo follow their own workflows. Test the exact model settings and account tier before committing.
| Feature | MiniMax H3 | Hailuo 2.3 | Sora 2 | Veo |
|---|---|---|---|---|
| Native Audio | Available | Available | Available | Available |
| Clip length | 4–15s | Varies | Varies | Varies |
| Native Stereo Audio | Available | Provider dependent | Available | Available |
| Omni-Reference | Supported | Limited | Provider dependent | Supported |
| Instruction Editing | Supported | Limited | Provider dependent | Supported |
| Image Reference Input | Available | Available | Provider dependent | Available |
| Prompt Capacity | Up to 7,000 chars | Provider dependent | Provider dependent | Provider dependent |
| Resolution Claim | Up to 2K | Varies | Varies | Varies |
| API Access | Official API available | Available | Available | Available |
| Free Tier | Free to start | Provider dependent | Varies | Varies |
Choose by production need, not a single headline spec. MiniMax H3 now has an official API with confirmed 768P and 2K output, 4–15 second duration, and multimodal reference inputs. Other capability comparisons remain provider-dependent, so confirm current settings and pricing before committing a budget.
MiniMax H3 Cinematic Scenes
Five cinematic scenes created by Alex Patrascu with MiniMax H3 demonstrate 15-second shots, dialogue, camera direction, and native audio. Play the full collection to compare storytelling and consistency across prompts.
Creator example: Alex Patrascu (@maxescu) · Source: XMiniMax H3 Pricing
Start with free credits, then choose a subscription or one-time credit pack when you need more video generations. Every option shows its credit allowance before checkout, so you can compare the likely cost of testing prompts, animating reference images, and producing finished clips without a long-term commitment.
Choose Subscription Plan
Basic
Perfect for individuals and light users
Billed annually at $90.00
Save 50%INCLUDES
- 300 credits / month
- ~22 seconds of AI video
- ~11 clips (2s, native audio)
Pro
For professional creators and teams
Billed annually at $234.00
Save 50%INCLUDES
- 1600 credits / month
- ~120 seconds of AI video
- ~59 clips (2s, native audio)
Max
Designed for large enterprises and professional studios
Billed annually at $960.00
Save 50%INCLUDES
- 9200 credits / month
- ~690 seconds of AI video
- ~340 clips (2s, native audio)
To cancel your subscription, please contact support: support@minimax3.com
MiniMax H3 FAQ
Plain answers about naming, current specifications, references, model differences, pricing, and this independent tool.
Start Creating with MiniMax H3
Begin with one clear scene, choose MiniMax H3 text-to-video or image-to-video, and judge the visuals and native audio together. Free starting credits let you test the workflow before choosing a larger plan. Check the generator settings for each run because available options can vary by mode.