Face
AI video that keeps your face
The sample person is not copied. Your photo is the face the model has to hold.
Seedance receives two images: your photo first, then a still from the sample. The instruction says to copy the camera and the transformation, and not the sample person’s face. H3 uses your photo as the first frame.
A clear, unobstructed face holds better than a tiny face in a wide shot. If the card says it wants a portrait, start with a portrait.