Inspiration
The inspiration for Pulse came from experimenting with the Gemini 1.5 Pro model and its long context window in Google AI Studio. Youtube comments have been an integral part of the experience of viewing videos, allowing both the creator and the viewer to connect with a larger community and gain insights and engage in fruitful discussion generated by the video. Reading comments for videos with a large amount of comments is tedious, and through the use of Gemini, can now be done much more efficiently and effectively.
What it does
Pulse efficiently scans all comments on a YouTube video, utilizing the advanced natural language processing capabilities of Google Gemini 1.5. This technology enables Pulse to generate comprehensive summaries that encapsulate the core themes and sentiments of viewer discussions. With this tool, the need for repetitive summarization using small context windows and outdated algorithms is eliminated. Pulse delivers highly accurate, detailed summaries informed by every comment, providing a complete overview of viewer feedback in one go.
Challenges we ran into
Long context prompts can be cumbersome to process, which hindered the user experience as analysis would take a considerable amount of time to be presented to the viewer. As such, we utilized a cloud database to store analysis that was completed for each video viewed, allowing for faster access.
Accomplishments that we're proud of
We're proud of building a tool that utilizes the cutting edge of technology to this date, and provides a useful and novel service to both creators and viewers alike on Youtube.
What we learned
Learning to build a Chrome extension effectively, as well as integrating with the Gemini and Youtube API's.
What's next for Pulse
The next step for Pulse is to focus on developing products to enhance the creator experience. Pulse aims to offer comment analysis services to creators to enable access to high-quality analysis and feedback for their content, and giving user's the opportunity to better engage with their communities. We will achieve this through flexible update schedules, allowing content creators to choose how frequently their video comments are summarized—from daily to weekly updates, depending on their preference. This flexibility helps in managing costs effectively, as each video is processed only as often as necessary. Once a video's comments are summarized, these summaries are stored and made available to subscribers extremely quickly, enhancing viewer engagement without redundant processing. Additionally, we hope to enable Pulse to asynchronously summarize comments across an entire video library, making it an ideal solution for creators with extensive content who wish to maintain engagement and understand viewer sentiment across all their videos.
Check out our Github README for more in-depth information.
Log in or sign up for Devpost to join the conversation.