MiniMax launches Hailuo S2V-01, a subject-reference video model for character consistency from one image
AI video model releases
MiniMax launches Hailuo S2V-01, a subject-reference video model for character consistency from one image
MiniMax released S2V-01, a feature of its Hailuo video model that keeps a character's face and identity consistent across a generated video from just one reference image.
The company claims S2V-01 is the first model to maintain character consistency in dynamic video generation and says it surpasses even fine-tuned models on this task.
Users can adjust posture, expressions, and lighting through text prompts while the model preserves the subject's identity across frames.
MiniMax acknowledges early-stage limitations: the model sometimes follows prompts less precisely than its T2V or I2V models and can show environmental morphing.
MiniMax says future updates will add support for multi-subject references, object references, and more complex multi-layered scenes.
cho.sh