EchoMimic is an advanced face image animation model capable of generating realistic facial videos driven by audio and selected facial keypoints, either individually or in combination. Through innovative training strategies, it addresses the instability potentially encountered in traditional audio-driven methods and the unnatural results that may arise from keypoint-driven approaches. EchoMimic has been comprehensively evaluated on multiple public and self-collected datasets, demonstrating superior performance in both quantitative and qualitative assessments.