Higgsfield Soul Id

This trains a persistent face model that Higgsfield can reuse across image and video generations instead of doing one-off face swaps. You feed it 5 to 20 photos of someone, wait a few minutes, and get back a reference ID that works with models like text2image_soul_v2 or soul_cinema_studio. It's meant for cases where you want the same person appearing consistently in generated content, like building a digital twin or avatar. Requires a paid plan. The skill handles the full loop: photo upload, training submission, polling until done, then handing you the ID to pass into higgsfield-generate. Straightforward if you need identity-faithful outputs and don't want to re-upload photos every time.