Each character must replicate exactly the same facial expression from the input images prompts

very few results