Background: Epilepsy is a very invalidating pathology characterized by the unpredictable appearance of abnormal cerebral activity leading to seizures and co-morbidities. The ability to detect and even predict seizures is a major challenge and many research laboratories are using rodents' models of epilepsy to unravel possible mechanisms. The gold standard to record and detect seizures is electroencephalography, but it is very invasive. For rodents used in research, video analysis is a very interesting approach but the major disadvantages are that it is time consuming, prone to human error, and not very reproducible. Commercial solutions for detailed phenotyping analysis on humans or rodents exist but they are costly. Some open source software programs are also available, they provide very interesting and precise behavior data, but none of them are made for high throughput analysis of a large number of video files generated by long lasting recordings. New method: We developed an open-source python-based package of two software programs that enable automated video acquisition and simple motion analysis associated with a spectral power analysis, which enable a semi-automated identification of convulsive seizures. The method needs cheap webcams and a computer or a server. Results: Using two murine epilepsy models (Nav1.1 mutations), we have compared our motion analysis software to human visual inspection and found an 88.8% accuracy in convulsive seizures detection. We then compare our method to the gold standard electrocorticogram analysis and found a 93.2% accuracy. The motion analysis is also interesting to get a readout of the animal activity without the invasiveness of electromyogram recordings. Conclusions: This new method is easy to use, cost-effective and allows: 1) detection of convulsive seizures in a noninvasive way, 2) high speed analysis of a large number of video files with a good accuracy, and 3) automated acquisition and semi-automated analysis of a very large number of files.