Summarizing full transcripts from lengthy speech-based videos using large language models is computationally expensive, resource-intensive, and requires further review to identify the most relevant sections. To address this, we propose an interactive web-based system for efficient textual summarization of speech-based videos, enabling on-demand, selective summarization of transcript segments. Our approach consists of multiple stages: first, the speech transcript from the video is extracted and chunked into segments. Next, representative topics are assigned to each segment along with informative visual cues, allowing users to filter segments of interest. Finally, textual summaries of the speech corresponding to the desired segments can be obtained on-demand using a local Large Language Model (LLM). This system allows users to navigate and selectively choose parts of videos they are interested in summarizing, reducing both inference time and computational costs. A video demonstrating the system in action is available at the following link: https://drive.google.com/file/d/1L7OPbkpnVFIagM6zIQwdTviOjxQ3ptSt .