Hailuo AI
MiniMax video generator: 2K native-audio clips, camera commands, multi-reference input
Hailuo AI is MiniMax's consumer video and image app. Its current flagship is H3, also marketed as Hailuo 3.0, released 31 July 2026, which generates up to 15-second clips at 2K with synchronized stereo audio and accepts up to 12 mixed image, video and audio references. Earlier Hailuo 2.3 and 02 models remain available.
- 16 creators using
- 1 news item tracked
- 2023 first launched
- Freemium pricing
- Moderate learning curve
What Hailuo AI can do
- Infinite canvas Not confirmed
- AI agent Yes
- Node / flow builder Not confirmed
- Public API Yes
- MCP server No
14 of 26 capabilities confirmed
| Capability | Supported | Details |
|---|---|---|
| Text to image | Yes | Create Image mode on the home page |
| Image editing | Not confirmed | |
| Text to video | Yes | H3 accepts text alone as input |
| Image to video | Yes | Start/end frame and image reference input |
| Video to video | Yes | V2V motion transfer on H3 |
| Native audio in video | Yes | H3 generates synchronized stereo audio with dialogue, effects and music |
| Lip-sync | Not confirmed | |
| Voice / TTS | Not confirmed | |
| Music & SFX | Partial | Music and SFX are generated jointly with H3 video; no standalone music tool in the Hailuo app |
| Upscaling | Not confirmed | |
| Character consistency | Partial | Omni Reference takes up to 9 image references to hold character identity |
| Camera control | Partial | Bracketed camera commands ([Push in], [Pan left], [Tracking shot]) documented for Hailuo 2.3 and 02 |
| Custom training | Not confirmed | |
| Third-party models | Not confirmed |
| Capability | Supported | Details |
|---|---|---|
| Infinite canvas | Not confirmed | |
| Node / flow builder | Not confirmed | |
| AI agent | Yes | Media Agent (formerly Hailuo Video Agent) plans and runs multi-step creation from a brief; a Hailuo Agent tab is on the home page |
| Chat assistant | Not confirmed | |
| Team collaboration | Not confirmed |
| Capability | Supported | Details |
|---|---|---|
| Public API | Yes | MiniMax Open Platform video API; H3 also listed on third-party routers |
| MCP server | No | No official MCP server |
| Mobile app | Yes | Hailuo mobile apps, updated August 2026 |
| Desktop / local | No | Cloud service |
| Open source | Partial | MiniMax said H3 open weights would follow |
| Capability | Supported | Details |
|---|---|---|
| Free tier | Yes | Free tier with limited credits and watermarked downloads |
| Commercial use | Yes | Paid-plan users keep commercial rights |
Yes Partial No Not confirmed by the vendor
What makes Hailuo AI different
- 01
Native stereo audio (dialogue, SFX, music) in the same pass as the picture, which removes a separate sound step for short clips
- 02
Up to 12 mixed image, video and audio references in one generation, more reference types than most single-prompt video tools
- 03
2K output on H3, above the 1080p ceiling of the Hailuo 2.x line
- 04
Bracketed camera commands for shot control, documented for the 2.3 and 02 models
- 05
Media Agent picks the model and steps from a brief, which helps producers who do not want to choose between models
About Hailuo AI
Hailuo AI is the consumer front end for MiniMax's video models. The headline change is H3 (Hailuo 3.0), launched 31 July 2026 as a general-purpose multimodal model rather than a narrow video revision. MiniMax's own announcement describes native 2K output at 24 fps, clips up to 15 seconds, joint modelling of voice, sound effects and music, native multi-shot planning, and video-to-video motion transfer. The reference system ('Omni Reference') accepts up to 9 images, 3 video clips and 3 audio clips, capped at 12 files in total, so a filmmaker can feed a character sheet, a motion reference and a voice track into one generation.
The Hailuo app now routes work through Media Agent (formerly the Hailuo Video Agent). You describe a finished piece and the agent picks models and steps, or you can run the pipeline yourself with uploaded media. Pre-built templates, an asset library and an image mode sit alongside it. Director-style camera commands in brackets such as [Push in] and [Pan left] are documented in MiniMax's API for the 2.3 and 02 models.
Hailuo fits a film workflow best as a fast previs and B-roll generator, a source of stylised anime or illustration shots, and a cheap way to test a shot before committing to a heavier model. Its strengths are the native audio, the multi-reference input and the price per second relative to the mainstream US models MiniMax compares itself against. It is weaker on node-based workflows, a canvas workspace and team review.
What's new in Hailuo AI
-
H3 Max becomes the recommended default
MiniMax H3 Max (Hailuo 3.0) is the default, with Hailuo 2.3 and 2.0 still available.
Source: help.scenario.com ↗ -
H3 (Hailuo 3.0) launches
General-purpose multimodal video model: up to 15s at native 2K, synchronized stereo audio, multi-shot modelling, and V2V motion transfer.
Source: minimax.io ↗ -
Omni Reference accepts up to 12 mixed references
Up to 9 images, 3 video clips and 3 audio clips per generation, with each video or audio reference 2 to 15 seconds and 15 seconds of combined reference duration.
Source: rits.shanghai.nyu.edu ↗
Models available in Hailuo AI
-
Hailuo 2.3 Latest MiniMax Standard and Fast variants; 768p and 1080p tiers; no last-frame conditioning Also: Hailuo 02
-
Key Features
- H3 (Hailuo 3.0) text, image, video and audio input in one generation
- Native stereo audio with dialogue, sound effects and music
- Omni Reference: up to 9 image, 3 video and 3 audio references per generation
- Video-to-video motion transfer
- Media Agent that selects models and steps from a brief
- Bracketed camera commands for shot control (documented for 2.3 and 02)
- Create Image mode alongside video
- Pre-built video templates and an asset library
- English and Chinese prompt support
- Mobile apps alongside the web app
When to reach for it — and when to skip
Reach for it when…
- Native audio in the same generation, so short clips arrive with sound
- Reference inputs cover characters, motion and voice in one pass
- 2K output on the current flagship
- Stylised and anime output is a known strength of the Hailuo line
- Agent mode lowers the cost of choosing between models
Skip it when…
- Free tier is watermarked and small
- Clips top out at 15 seconds per generation
- No canvas, node-workflow or team review features
Best For
✓ Ideal for
- Previs and animatics that need native sound in the first pass
- Stylised anime and illustration shots for music videos and short films
- Character shots driven by a reference image plus a motion or voice reference
- Quick B-roll generation for explainer and documentary edits
- Test shots before committing budget to a heavier model
✗ Not built for
- Long-form continuous shots beyond 15 seconds without stitching
- Node-based or canvas-led pipelines
- Teams that need built-in review and approval workflows
Working Tips from Filmmakers Using Hailuo AI
- 01 Use the Omni Reference slots deliberately: one character image, one motion clip and one voice clip usually gives more control than a long prompt
- 02 Keep each reference clip between 2 and 15 seconds and the combined reference length at or under 15 seconds, or the generation will reject it
- 03 Generate audio-first shots for dialogue scenes, since native audio fixes lip timing in the first pass
- 04 Check the watermark status on the free tier before you use a clip in a client deliverable
Hailuo AI pricing in 2026
| Plan | Price | What you get |
|---|---|---|
| Free | $0 monthly |
|
Prices checked against Hailuo AI's pricing page on 9 Oct 2026. Vendors change plans often, so confirm before you buy.
Hailuo AI vs Runway vs Kling AI vs Higgsfield
| Hailuo AI | Runway | Kling AI | Higgsfield | |
|---|---|---|---|---|
| Price | Free tier (limited credits, watermarked); paid plans: see the vendor's pricing page | Free (125 one-time credits); Standard $15/mo ($12/mo billed annually) | See the vendor's pricing page | Free tier; paid plans from €19/mo billed annually (prices exclude VAT) |
| Infinite canvas | Not confirmed | Not confirmed | Partial | Partial |
| Node / flow builder | Not confirmed | Yes | Not confirmed | Not confirmed |
| AI agent | Yes | Yes | Not confirmed | Yes |
| Chat assistant | Not confirmed | Yes | Not confirmed | Yes |
| Team collaboration | Not confirmed | Yes | Not confirmed | Yes |
| Text to image | Yes | Yes | Yes | Yes |
| Image editing | Not confirmed | Yes | Not confirmed | Yes |
| Text to video | Yes | Yes | Yes | Yes |
| Image to video | Yes | Yes | Yes | Yes |
| Video to video | Yes | Yes | Yes | Yes |
| Native audio in video | Yes | Partial | Yes | Yes |
| Lip-sync | Not confirmed | Not confirmed | Yes | Not confirmed |
| Voice / TTS | Not confirmed | Yes | Partial | Yes |
| Music & SFX | Partial | Yes | Not confirmed | Yes |
| Upscaling | Not confirmed | Yes | Not confirmed | Yes |
| Character consistency | Partial | Partial | Yes | Yes |
| Camera control | Partial | Not confirmed | Yes | Yes |
| Custom training | Not confirmed | Partial | Not confirmed | Not confirmed |
| Third-party models | Not confirmed | Yes | No | Yes |
| Public API | Yes | Yes | Yes | Yes |
| MCP server | No | Yes | Not confirmed | Yes |
| Mobile app | Yes | Yes | Not confirmed | Partial |
| Open source | Partial | No | No | No |
| Free tier | Yes | Yes | Yes | Yes |
| Commercial use | Yes | Not confirmed | Yes | Not confirmed |
Each column comes from that tool's own researched page. "Not confirmed" means the vendor doesn't say.
Open the full Hailuo AI vs Runway vs Kling AI vs Higgsfield comparison →Use Cases
Integrations
Tags
Creators Using Hailuo AI
Ai Filmmaker • Cinematic Storyteller • Short Film Creator
AI Filmmaker - Cinematic Storyteller - Google Flow - Seedance - Pixverse - Kling - Sora - Runway - Hailuo - NanoBanana - GPT - Flux - Gemini
AI FILMAKER
AI Filmmaker & Creative Produce at Rawi AI Studio
Sources
Facts on this page were checked against Hailuo AI's own pricing, product and changelog pages on , then cross-checked with independent coverage where vendor pages were silent. Anything we couldn't confirm is marked "Not confirmed" rather than guessed.
- MiniMax H3 announcement (31 Jul 2026) minimax.io
- Hailuo AI home page (H3 live, modes, Hailuo Agent) hailuoai.video
- NYU Shanghai RITS: H3 release and reference limits rits.shanghai.nyu.edu
- Scenario help: MiniMax Hailuo video (updated 2 Sep 2026) help.scenario.com
- Runway pricing (comparison column) runway.com
- Hailuo AI Pricing 2026 aiarty.com
Spotted something out of date? Tell us and we'll re-check it.
Discussion
No comments yet — be the first.