Recently loaded
Download
1 variant available
fp32 SafeTensor
mila_v1_h3_video_gl_rank32_v1-totalstep00002000.safetensors
Full precision, largest file • 568.82 MB
Verified: 5 days ago
Stats
1540 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
130 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
8170 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
Reviews
(9)
Published
Oct 20, 2026
Base Model
MiniMax H3
Training
Steps: 2,000
Epochs: 11
Usage Tips
Strength: 1
Hash
AutoV2
2D93E3E040
Trigger Words
mila_v1
MiniMax H3 release of the Mila character LoRA, trained for consistent identity in photorealistic video generation.
The MiniMax H3 Community License Agreement’s grant excludes the European Union, the United Kingdom, the Republic of Korea and the United States of America.
MiniMax H3
Mila is a fictional character LoRA designed for consistent generation across LTX-2.3/2.5 (T2V-I2V), MiniMax H3 (T2V-I2V) as well as Z-Image Turbo and Krea 2.
Mila works especially well for realistic lifestyle content, fashion scenes, luxury environments, social media content, outdoors, fitness, travel, and cinematic character shots.
Triggers:
• MiniMax H3: mila_v1
• LTX-2.3: mila or mila_v1
• Z-Image Turbo: mila_v1
• Krea-2 Turbo: mila_v1
Recommended use cases:
• Consistent fictional character generation
• Realistic character video generation
• Portrait and lifestyle photography
• Fashion, luxury, fitness, travel, and social media content
• Luxury hotel interior and exterior cinematic scenes
• Stable facial identity across different outfits, poses, and locations
• Fashion Advertising