What problem does it solve? Creating consistent, identity-faithful images and videos of a real person with generative AI requires a trained identity model, and the training workflow involves CLI setup, authentication, photo selection, and long-running job polling that is easy to get wrong. ## Core Features & Use Cases - Soul Character Training: Submits 5-20 face photos to Higgsfield to train a reusable identity model, returning a reference ID for all future generations. - Variant Selection: Chooses between the image-focused soul-2 variant and the cinematic soul-cinematic variant based on the intended downstream use. - Guided Photo & Troubleshooting References: Includes a photo guide covering quantity, angles, lighting, and quality, plus troubleshooting for plan limits, training failures, and expired sessions. - Use Case: A creator wants their own face in AI-generated cinematic videos. They provide a name and ten varied photos; the Skill trains the Soul, waits for completion, and hands back a reference ID usable with higgsfield-generate via --soul-id. ## Quick Start Train a Soul character named alex using the face photos in my photos folder so I can use my identity in Higgsfield generations.
