Temporal echo image registration is a basis for clinical quantifications such as cardiac motion estimation, myocardial strain assessments, and stroke volume quantifications. Deep learning image registration (DLIR) is consistently accurate, requires less computing effort, and has shown encouraging results in earlier applications. However, we propose that a greater focus on the warped moving image's anatomic plausibility and image quality can support robust DLIR performance. Further, past implementations have focused on adult echo, and there is an absence of DLIR implementations for fetal echo. We propose a framework combining three strategies for DLIR for both fetal and adult echo: (1) an anatomic shape-encoded loss to preserve physiological myocardial and left ventricular anatomical topologies in warped images; (2) a data-driven loss that is trained adversarially to preserve good image texture features in warped images; and (3) a multi-scale training scheme of a data-driven and anatomically constrained algorithm to improve accuracy. Our experiments show that the shape-encoded loss and the data-driven adversarial loss are strongly correlated to good anatomical topology and image textures, respectively. They improve different aspects of registration performance in a non-overlapping way, justifying their combination. We show that these strategies can provide excellent registration results in both adult and fetal echo using the publicly available CAMUS adult echo dataset and our private multi-demographic fetal echo dataset, despite fundamental distinctions between adult and fetal echo images. Our approach also outperforms traditional non-DL gold standard registration approaches, including Optical Flow and Elastix. Registration improvements could also be translated to more accurate and precise clinical quantification of cardiac ejection fraction, demonstrating a potential for translation.