I’ve been experiencing an issue where YT dataset URLs do not seem to update even after downloading the transcript. When I insert the URL and click on ‘Download Transcripts’ and ask the AI what the video is about, it always provides information from an old YT video transcript.
Video about the issue: https://youtu.be/zRnqXf-lVbo
Thank you for the post. Can you please try removing the old transcript by using the button in the Dataset panel to navigate to the folder where it is stored, then try and download the new transcript again?
its an annoying work around which does not work 100% of the time as well. I tried deleting the temp files and it still had it cached somewhere.
I agree that cannot be the fix to this issue. Having to go to the folder location and delete the files there just to download the transcripts for the video, the script is downloading the transcript in txt format using the YouTube video ID (7pSqk-XV2QM.txt). It can successfully download more transcripts without issue, but it always uses the first video you’ve downloaded as the dataset, even if you’ve just downloaded a new transcript. It can’t be the fix, as it’s annoying to the users and sometimes unreliable.
I have been hitting the same issue.
As a workaround, I have been downloading transcripts directly from YouTube then storing them in a separate directory and setting that as the local data source.
It works, but it is messy.