Readme

MiniMax-Hailuo-2.3

MiniMax-Hailuo-2.3 is a video generation model family designed for realistic motion, improved visual consistency, and high-fidelity stylization. It supports both text-to-video and image-to-video workflows, with two performance profiles: a standard model with broader capabilities, and a fast model variant optimized for reduced latency.

Model Variants

MiniMax-Hailuo-2.3

The core model supports both text and image as input. It is intended for cinematic workflows, realistic animation, and high-detail scenes.

• Input: Text and Image
• Resolution: 768p and 1080p
  1080p videos are limited to 6-second duration
• Duration Options: 6 seconds or 10 seconds
• Aspect Ratio
• Image-to-Video: follows source image
• Text-to-Video: defaults to 16:9
• Last Frame Handling: not supported

This model provides the highest visual quality and the broadest feature coverage within the family.

MiniMax-Hailuo-2.3-Fast

The fast variant is optimized for turnaround time and computational efficiency.

• Input: Image only
• Resolution: 768p and 1080p
  1080p videos are limited to 6-second duration
• Duration Options: 6 seconds or 10 seconds
• Aspect Ratio
• Image-to-Video: follows source image
• Last Frame Handling: not supported

This version offers quicker feedback during iteration or prototyping cycles.