For what you're trying to do, I'd recommend image-to-image generation from a single 3d model. Get familiar enough with SFM to make your fake person, take screenshots, then have SD run through them with the right level of freedom to launder the 3d models into something more photorealistic.
That should keep the "individual" consistent. It won't be quick, and it won't be easy, but unless you're telling SD to steal the likeness of a famous person it already has reference art from in its model, it won't be able to keep images to a single "person".
For what you're trying to do, I'd recommend image-to-image generation from a single 3d model. Get familiar enough with SFM to make your fake person, take screenshots, then have SD run through them with the right level of freedom to launder the 3d models into something more photorealistic.
That should keep the "individual" consistent. It won't be quick, and it won't be easy, but unless you're telling SD to steal the likeness of a famous person it already has reference art from in its model, it won't be able to keep images to a single "person".