Visual Quality and Transferability are the two key property of adversarial face examples. However, few works consider to improve the key properties simultaneously. To this end, we proposed a novel adversarial attack called Adversarial Restoration (AdvRestore) that enhances the visual quality and the transferability of adversarial face examples simultaneously by using a face Restoration Latent Diffusion Model Prior. Specifically, we first train a face Restoration Latent Diffusion Model (RLDM) for face restoration. Then, we use RLDM to restore of the attacker image, in the process of restoration, we add adversarial perturbations on the output feature of the UNet of RLDM. Combined with the prior, the visual quality and tranferability of the crafted adversarial face examples can be further improved. Experimental results demonstrate the effectiveness of our proposed attack method.