Generalization of countermeasure (CM) systems is crucial for voice anti-spoofing in real-world scenarios. To that effect, in this paper, an attempt is made towards the generalization of CM system through the cross-database evaluation between the standard statistically meaningful corpora, namely, ASVSpoof 2017 V2.0, ASVSpoof 2019, and Voice Spoofing Detection Corpus (VSDC). To the best of authors’ knowledge, VSDC is the only dataset that which contains both One-Point Replay (1PR) and Two-Point Replay (2PR) utterances. Furthermore, we have investigated the effect of reverberation and noise suppression capabilities of TEO on the different types of replay signals, i.e., 1PR and 2PR signals. To that effect, we have compared performance of different variants of TECC w.r.t frequency scale (i.e., Linear, Mel, and Inv-Mel) w.r.t state-of-the-art features, such as MFCC, CQCC, LFCC. TECC is found to outperform the rest of the features and thus, validating our hypothesis of capturing reverberation more efficiently via TECC. TECC (Mel) gave % EER of 31.65% and 31.96% with training on ASVSpoof 2017 and testing on 0PR-1PR and 0PR-2PR VSDC, respectively. Further, % EER for TECC (Linear) with training on VSDC and testing on ASVSpoof 2017 for development and evaluation sets are 20.01% and 27.63%, respectively. Finally, the analysis of latency period is presented where TECC gave the optimal performance, indicating the potential for practical Spoofed Speech Detection (SSD) system development.