Inspiration
Videos account for over half of internet data traffic, with 53.72% of all global data traffic attributed to video content. Processing transcripts of these videos can have multiple applications. Conducting sentiment analysis on videos, summarization of video content as well as using video transcripts to train LLMs to create chatbots are high-return solutions that are dependent on video transcription.
What it does
This project creates transcripts of videos present in a local directory and stores them in a .txt file at a given path. This project is made for the Hackatra 2024 hackathon.
How we built it
Algorithm Outline: Traverse Local Directory: Implements a script to recursively search and identify video files within a specified directory structure. Convert Videos to Audio: Utilizes ffmpeg to extract audio tracks from identified video files, ensuring compatibility and quality. Transcription Using Whisper: Integrates Whisper for local audio transcription, leveraging its accuracy and efficiency in converting speech to text. Cooling Time Implementation: Incorporate intervals between transcription tasks to prevent system overheating, enhancing long-term stability. Generates Combined Transcript: Merges individual transcriptions into a cohesive transcript document, ensuring clarity and coherence.
Challenges we ran into
This was my first time working with ffmpeg and whisper. It took me some time to learn how to work with them.
What we learned
I learn how to convert video files to audio and audio to text.
What's next for Video Transcription
Moving forward this application can be extended to train llms to create chatbots, generate minutes of the meetings and lecture notes
Log in or sign up for Devpost to join the conversation.