Hailuo 02

Hailuo 02 Description

MiniMax's Hailuo 02 is a cinematic AI video model engineered for high-fidelity content creation and ranked #2 globally on the Artificial Analysis benchmark. Building on a diffusion-transformer architecture, it delivers photorealistic visuals, advanced physics simulation, and director-level controls for professional-grade video production.

‍

‍Technical Specification

Performance Benchmarks

Hailuo 02 is optimized for cinematic realism with advanced rendering and user control.

• Video Resolution: Full HD (1920x1080p) native output.
• Video Length: Up to 10 seconds per generation.
• Generation Speed: 30 seconds to 5 minutes, depending on prompt complexity.
• Performance Benchmark: #2 global rank on Artificial Analysis with a 1332 ELO score, outperforming Google's Veo 3.

• API Pricing:
  
  • 768P 6s: $0.364;
  • 768P 10s: $0.728;
  • 1080P 6s: $0.637

Performance Metrics

In formal evaluations, Hailuo 02 has achieved high rankings. It secured the #2 global position on the Artificial Analysis benchmark for image-to-video generation. This places it ahead of Google's Veo 3 and just behind Seedance 1.0 from ByteDance. User benchmarks and head-to-head comparisons also suggest Hailuo 02 is a strong contender, with some users claiming it is superior to Veo 3 in performance and realism. While Google's Veo is noted for its motion fluidity, users have reported that Hailuo 02 often produces more cinematic emotion.

Key Capabilities

Hailuo 02 delivers precise outputs for complex video workflows with a suite of professional tools.

• Director Control Toolkit: Supports text-based commands for cinematic camera movements like "pan-down," "dolly zoom," and "bird's eye view".
• Advanced Physics Simulation: Excels at rendering realistic environmental effects, including water, fog, light, and material interactions.
• Character Consistency: Maintains character identity and appearance across shots using subject referencing and body tracking technology.
• Multi-Format Input: Creates video from both text prompts and reference images.
• High-Fidelity Rendering: Produces videos with impressive frame-to-frame consistency and minimal visual distortion.
• First/Last Frame to Video: Ensures professional scene transitions and offers “camera language” parameters for a more cinematic result.

‍

Optimal Use Cases

• Social Media Content: Creating engaging short-form videos for platforms like YouTube Shorts, TikTok, and Instagram Reels.
• Marketing and Branding: Producing product teasers, stylized promotional visuals, and branding montages without a full production crew.
• Creative Prototyping: Developing visual storyboards, art film vignettes, and music visualizers.
• Narrative Storytelling: Building coherent multi-shot narratives, particularly in genres like urban dystopia, fantasy, and surrealism.

‍

Code Samples

Video Generation

‍

Get Generated Video

‍

Parameters

• "model": string
• "prompt": string - The text description of the scene, subject, or action to generate in the video.
• "duration": 6 | 10 - The length of the output video in seconds.
• "resolution" : "768P" | "1080P" - The dimensions of the video display. 1080p corresponds to 1920 x 1080 pixels, 768p corresponds to 1366 x 768 pixels.
• "prompt_optimizer": boolean - If True, the incoming prompt will be automatically optimized to improve generation quality when needed. For more precise control, set it to False — the model will then follow the instructions more strictly.
• "first_frame_image": url - The model will use the image passed in this parameter as the first frame to generate a video

‍

Comparison with Other Models

• Vs. Google Veo 3: Ranked higher on the Artificial Analysis benchmark (#2 vs. #3), with users noting Hailuo 02 produces more cinematic emotion while Veo excels in motion fluidity.
• Vs. Hailuo 01: A significant upgrade offering higher 1080p resolution, superior character consistency, and more dynamic shot composition.
• Vs. Seedance 1.0: Positioned just behind the top-ranked model, making it a leading contender in the market

API Integration

Accessible via AI/ML API. Documentation: available here.