Due to the rapid development of face forgery technology, the corresponding face forgery detection methods are constantly facing challenges. In this paper, we utilize multi-feature consistency of images to address the face forgery detection problem and propose a multi-feature consistency learning (MFCL) method for this task. Specifically, considering of the multi-feature consistency of face images, our MFCL method extracts different types of features from each image, leverages different feature representations of the face image during the training stage, generates prediction regions for multiple models independently, and fuses the information of the models under different features to guide the training, allowing it to learn the composite features of the face image, thus improving the accuracy and generalization ability.