The Best approach is to get a set of let's say 12-20 images with PulID , edit expression and maybe posture with open pose ( I personally use plain kolors with pulid and many text2image prompts.I get better results than with face swap or other models) and make your character in different pictures. Try that some of them cover also the full body aspect. Later Use all this images to train a Lora..there is no way you can skip this process if you wanna make a consistent character in all your future pictures