What projects are you working on

[ABearden]

I've been "following," and I'll use the word loosely cause its not an indepth tutorial but it's a video on YouTube by kalle hallden about making a Google for Joe Rogan. https://youtu.be/UUnAcrzA0nA

Basically everything he's done in that video is my only point of reference as to how to build this project.

Also, your code example would work perfectly for hyperlinking to specific sentences, my only issue is writing the algorithm that would format the vtt file as such. Unless I can skip any conversion process all together and simply have the raw vtt files be searchable. I'm just unsure of how to get the same formatting that a JSON file can give, but to get that with vtt or txt. That's why the conversion seems important to me. But like I said, it's a little out of my wheelhouse which isn't very large in the first place lol

JSON, VTT, the database won't care: it's all text. Some database engines are JSON aware and allow for access to the contained structure, but that's typically slower than raw table access. But if you're storing a bunch of text files in a column, "WHERE transcription LIKE '%UFO%' " will still return all the rows that contain the search text regardless of what format it's in. You could just dump each video into the table with columns like ID, url, name, and transcription and get something super basic going. If you have a mixture of JSON and VTT files, you could even do two columns: fileType and fileContents. Then you just search the file contents and any further processing can reference the fileType so it knows what to expect.

There's a lot you can do with pre-processing the text to get word scores and rank by relevance, but that's the next tier up of development effort.