Yu2bias 1.19 beta

Summarize YouTube videos and analyze them for bias using ChatGPT

Still early days so errors and glitches will happen, I appreciate your patience! Summarizing the comments takes time - up to 30 seconds. When user has a YouTube video with the transcript open, this will summarize, analyze bias, and provide counterpoints. Fact check! Logical fallacy check! Custom instructions! Comments section summary! .. and transcript extraction are available! ** Warnings about extension being able to read and write all your YouTube site data: This extension requires permission to see the video's title, duration and to get transcript. The extension requires permission to write the results of the summary to a new tab. ** ** No search data, browser history, special intsructions are stored, no analytics are collected. Not on the backend server, nor any device ** ** In the future I may implement a optional survey of custom instructions to help fine tune this app. Users will be notified when and if, and users will be able to opt out ** Nerd stuff: The transcript is scraped from the tab (no automated way is possible rn, currently google is blocking all ips that use these tools) The video ID is taken from the open tab, and that is sent to an AWS Lambda function which gets the video's title and duration from the YouTube Data API. The transcript is sent to the ChatGPT4o-mini API via an AWS Lambda function. The max_token is set to 4096 which seems to be good for a 60minute video transcript. ChatGPT is instructed to summarize, provide bias analysis, etc and return results. Chat's Google search response is parsed for the query, and that query is sent to a Google custom search API via AWS Lambda function. Final summary with counterpoint URLs is then parsed for the results tab. If user requests comments summary, the latest 1000 comments + replies are fetched via Youtube Data API via AWS Lambda function. These are then parsed and passed to ChatGPT for analysis. This process may add up to 30 seconds! The results are appended to the results page. If the user requests transcript extraction, that is scraped from the transcript window, and appended onto the results page.