Autonomic dysfunction in epilepsy is well-described. Heart rate variability (HRV) is a useful method to evaluate autonomic cardiac tone. Cardiac dysfunction may be involved in sudden unexpected death in epilepsy (SUDEP). HRV is a promising biomarker to enlighten the heart-brain axis role in SUDEP, but the required duration for a proper HRV recording in clinical routine remains unknown. This study aimed to verify the reliability of ultra-short HRV indices to evaluate cardiac autonomic tone in patients with epilepsy (PWE). Thirty-nine patients with mesial temporal lobe epilepsy (MTLE) had electrocardiogram recordings during the first day of video-EEG. Pearson\'s correlations were performed to evaluate the association between ultra-short HRV indices (five 1-min and five 30-s epochs) with standard time recording (5-min) and ANOVA compared the differences between mean HRV indices across epochs. Time domain (TD) indices showed higher mean r values when compared to frequency domain (FD) indices in 1-min (TD: r 0.80-0.99, FD: r 0.61-0.95) and 30-s epochs (TD: r 0.69-0.99, only high frequency: mean r values of 0.96). ANOVA evidenced that standard deviation of RR intervals and very low frequency means had at least 3 epochs significantly different for 1-min and 30-s epochs. Root mean square of the successive differences of RR intervals (rMSSD) presented higher Pearson\'s coefficient values and lower percentage of variation at 1-min or 30-s epochs in comparison to other HRV indices. In conclusion, rMSSD is the most reliable ultra-short HRV index for cardiac autonomic tone assessment in MTLE. The prognostic value of ultra-short HRV for cardiovascular risk evaluation in epilepsy remains to be determined in future studies.