Hailuo 2.3

byMiniMax

Generate fluid human motion, micro-expressions, and physical camera sweeps with locked lighting and zero mechanical jitter

Direct your scene

How Hailuo 2.3 works

From static frames or pure directorial prompts to full cinematic video in three deliberate moves.

Hailuo 2.3 text prompt connected to its native rooftop stunt video result in Fuser

Frame your opening shot

Upload a high-resolution starting image for frame-accurate animation, or begin with a pure text prompt.

Hailuo 2.3 node showing the complete header, sockets, Standard model, six-second duration, prompt optimization and native video result

Direct motion and optics

Add bracketed camera commands like [Pan left] and specify physical details such as fabric inertia or liquid splashes.

Hailuo 2.3 node displaying the native blue-hour rooftop stunt video result

Select tier and render

Choose Fast variants for rapid 768p layout drafting or flagship Pro tiers for uncompromised 1080p cinematic fidelity.

What Hailuo 2.3 is good at

Engineered with true physical inertia, native director camera framing, and micro-expression fidelity that keeps subjects grounded.

True Physical Inertia and Lighting

Simulate natural gravity, cloth inertia, and specular reflections that stay locked across high-speed dynamic camera movements without mechanical jitter.

Native Director Camera Directives

Control focal dynamics using bracketed directives such as [Dolly zoom] and [Push in] without relying on vague natural language workarounds.

Facial Micro-Expressions

Render subtle muscle cues, shifting gazes, and slight brow twitches in human close-ups while locking character identity across the entire clip.

Dual-Speed Production Tiers

Draft motion concepts instantly at 768p using the Fast image-to-video variant before committing credits to flagship 1080p Pro and Standard 6 or 10-second renders.

Made with Hailuo 2.3

Explore how physical lighting, garment dynamics, and fluid camera sweeps perform across photorealistic and stylized video renders.

Resonant stillness and material tension

High-velocity physical momentum

Thermodynamic particle and steam simulation

Cloth inertia and flowing drapery dynamics

Tactile surface physics and macro lighting

What people build with Hailuo 2.3

How filmmakers, fashion directors, commercial VFX artists, and animators harness physics-grounded video generation.

Cinematic Pre-Visualization

01

Test complex multi-axis camera movements and physical character interactions before arriving on set, validating shot blocking with true camera physics.

High-Fashion Lookbooks

02

Bring static editorial portraits to life with accurate garment inertia, silk drape physics, and lifelike character micro-expressions.

Commercial B-Roll Production

03

Generate dynamic liquid splashes, glass shattering, and dynamic product hero shots with flawless frame-to-frame lighting consistency.

Stylized Anime Animation

04

Produce fluid 2D Japanese anime, ink-wash sequences, and game CG with clean linework and persistent character identity across every frame.

Rapid Storyboard Prototyping

05

Leverage Fast image-to-video tiers to turn static storyboards into 768p animated animatics in seconds, dramatically accelerating visual sign-off.

Hailuo 2.3 specs

Plans and credits
Versions
Pro Fast, Standard Fast, Standard, Pro
Inputs
Prompt (text), First Frame (image)
Output
video
Duration
6 seconds, 10 seconds
Cost
265–772 credits per run

Frequently Asked Questions

The Pro variant delivers native 1080p resolution with maximum sampling depth for final cinematic production, whereas the Fast variant generates 768p video at significantly higher speeds for rapid drafting. Pro supports both text-to-video and image-to-video workflows, while Fast is strictly dedicated to image-to-video iteration at lower computational cost.

Use the Pro Image-to-Video tier when maximum texture detail, 1080p clarity, and subtle micro-expressions are essential. For rapid storyboard exploration and checking motion trajectories before committing to final renders, the Fast Pro or Fast Standard Image-to-Video variants offer faster turnaround at 768p.

Direct camera motion by placing bracketed director keywords at the beginning or end of your prompt, such as [Pan left], [Push in], or [Dolly zoom]. You can combine up to three directives within a single bracket to specify complex spatial trajectories without confusing the model's action rendering.

The Standard variant supports either 6-second or 10-second video generations. The 10-second duration is best suited for ambient scenes and continuous camera sweeps, as shorter 6-second clips minimize the likelihood of motion drift during complex physical action.

The model is not designed for multi-stage sequential actions within a single generation, such as a character running, stopping, and sitting down in sequence. It also avoids fine typography, vector logo animation, and exact last-frame loop matching, as it focuses its physical simulation on continuous motion starting from the first frame.

Try Hailuo 2.3 on Fuser

Generate fluid human motion, micro-expressions, and physical camera sweeps with locked lighting and zero mechanical jitter

Direct your scene

Related models

Hailuo 2.3 docsAll models

MiniMax H3 Recast

MiniMax

MiniMax H3

MiniMax

Hailuo 02

MiniMax

Kling O3 Edit

Kuaishou Technology

Kling O3

Kuaishou (Kling AI)

Kling 3.0 Video

Kuaishou

Hailuo 2.3 in our guides

All articles