The prevalence of memory deficits among epilepsy patients is approximately 50%, reported as a major factor affecting the quality of life. In particular, individuals with temporal lobe epilepsy are increasingly identified with a deficit termed Accelerated Long-Term Forgetting. Monitoring and treatment of these deficits require efficient automated tools for repeated probing of memory functions with or without concurrent recording of the brain activities. We used a classic verbal memory free recall paradigm, in which subjects recall out-loud words after a distractor, assessing performance by transcribing and detecting correctly recalled words from subjects' vocalizations. Annotating the exact beginning and end of the vocalizations is crucial for memory-related brain activity analysis. Traditional approaches utilizing manual marking and labeling all vocalizations in the task is laborious, time-consuming, and prone to human error, especially in case of large datasets. To overcome these limitations we developed an automated transcription interface. Our interface is based on Whisper, fine-tuned with the CommonVoice dataset to ensure accurate transcription across various linguistic contexts, such as Polish or Czech. We assessed transcription accuracy with Word Error Rate (WER), achieving promising results with WER of 6.9% for the Polish and 9% for the Czech languages. The interface's performance, in speed and accuracy of word detection and annotation, surpassed standard manual transcription. These findings demonstrate robust and efficient transcription accuracy achieved for a set of challenging Slavic languages. Such automated transcription interfaces have a potential to streamline data preprocessing and thus enhance biomedical research and technologies for human computer interfaces.