The media-higgsfield-identity skill handles visual consistency across multi-shot AI generations within the Higgsfield ecosystem. It intelligently evaluates whether user prompts require a custom-trained Soul Character model (ideal for high-fidelity, single-person digital twins) or an instant Reference Element (supporting multi-subject, props, environments, and non-human assets). Built to safeguard user credits and privacy, the skill enforces strict consent and approval gates before uploading human faces or triggering irreversible training jobs, preventing costly configuration mistakes and ensuring seamless continuity in image and video creation.
主要功能
01Seamless cross-model integration for downstream image and cinematic video workflows
02Pre-flight credit and duplicate character checks to prevent unnecessary training costs
03Intelligent routing between trainable Soul models and zero-shot Reference Elements
04Multi-subject scene composition support using dynamic Element prompt injection
05Multi-stage privacy protection with mandatory human face consent verification
06315 GitHub stars
使用场景
01Composing multi-character interactions and group portraits without dedicated identity model training
02Injecting consistent commercial product references and branded props across diverse advertising scenes
03Training digital twin Soul characters from personal photos for recurring single-person shots