Show HN: Search Bar for YouTube Transcripts

[katella]

Hi HN!

Using Ctrl+F on YouTube Transcripts doesn't have the best UI/UX, so I built a chrome extension to inject a input search box that filters for any queries.

YouTube actually has this feature for a few videos (ex: https://https://www.youtube.com/watch?v=oAHbLRjF0vo) for years now. I'm not sure why they haven't rolled it out for all videos. I didn't want to wait.

The extension will only inject it's own search box if the video does not already have one on YouTube. I tried styling it to look similar to YouTube's design.

Indexing and querying rules follow the newest Lunr.js rules.

[tobeagram]

This is such a good idea that's simple and well executed. The number of times, I've had to sift through YouTube videos to find the bit that's relevant... awesome work!

Have you got any further plans to improve it, or just want to keep it simple?

[katella]

Some new features could be:

- add it to other sites that have videos

- improve search with AI.

Some performance updates:

- parse transcripts from request, not the html

- filter using overlay (inject my own html) instead of mutating the DOM (loops through all elements and changes display attribute)

[katella]

Yes, do you have any suggestions?

[Yisz]

Very cool. Does this also generate the transcripts or only add search for videos with existing transcripts?

[katella]

Adds search for videos with existing transcripts (including auto generated). All client-side.

A future feature could be generating transcripts for videos without one.

[katella]

Does anyone know why YouTube hasn't made this a full feature? Corporate bureaucracy?

[jknutson]

Perhaps it could be misaligned incentives? The primary variable YT optimizes for is time spent on the platform. Being able to instantly find and skip ahead to the part you wanted means you didn’t spend the extra time that you would’ve otherwise.

Granted, I can’t see this specific feature really moving the needle one way or the other. But, internally, they might apply some sort of lens to their decision making that asks “does this increase or decrease time spent and eyeballs monetized?”, and if that is the case then there is a real argument to be made that this feature would not meet the standard for a “good feature”, based off that criteria.

Probably more likely that their entire UI is just so deeply fine tuned and A-B tested that any individual change to it has a lot of friction, lest they push an update that decreases watch time by 0.1%.