This paper proposes a Conditional Adversarial Autoencoder (CAAE) for face age progression and regression. The CAAE learns a face manifold, allowing smooth age progression and regression while preserving personality. Unlike existing methods that require paired samples, the CAAE uses a generative model to directly produce images with desired age attributes from unlabeled data. The face is first mapped to a latent vector through a convolutional encoder, then projected to the face manifold conditional on age through a deconvolutional generator. Two adversarial networks are used to enforce photo-realistic generation. The CAAE is robust to variations in pose, expression, and occlusion. Experimental results show that the CAAE outperforms existing methods in generating realistic faces for both age progression and regression. The framework is flexible, requiring no labeled data for training or testing, and can be applied to various image generation tasks. The CAAE is also effective in cross-age recognition and face morphing. The proposed method is compared with state-of-the-art approaches and ground truth, demonstrating its effectiveness in generating realistic faces across different ages. The CAAE is also robust to variations in pose, expression, and occlusion, making it suitable for real-world applications.This paper proposes a Conditional Adversarial Autoencoder (CAAE) for face age progression and regression. The CAAE learns a face manifold, allowing smooth age progression and regression while preserving personality. Unlike existing methods that require paired samples, the CAAE uses a generative model to directly produce images with desired age attributes from unlabeled data. The face is first mapped to a latent vector through a convolutional encoder, then projected to the face manifold conditional on age through a deconvolutional generator. Two adversarial networks are used to enforce photo-realistic generation. The CAAE is robust to variations in pose, expression, and occlusion. Experimental results show that the CAAE outperforms existing methods in generating realistic faces for both age progression and regression. The framework is flexible, requiring no labeled data for training or testing, and can be applied to various image generation tasks. The CAAE is also effective in cross-age recognition and face morphing. The proposed method is compared with state-of-the-art approaches and ground truth, demonstrating its effectiveness in generating realistic faces across different ages. The CAAE is also robust to variations in pose, expression, and occlusion, making it suitable for real-world applications.