solemn grove
I put together a simple website for searching through past Neuro-sama stream transcripts. Any feedback or discussion is welcome. If there's enough interest and people find it useful I'll continue working on it.
https://libraryofladev.com/
Update: The archive is now complete with 422 videos transcribed from the official and unofficial VOD channels!
Migrated to https://discord.com/channels/574720535888396288/1337595628607242282
proven vortex
Recently I'm trying to find a way to quickly search for certain sentences in neuro's streams to make clips, and this website really comes in handy. Thank you for the hard work and the nice website!
By the way, what's the range of streams that are included? It seems results from the subathon are not shown. And will there be a fuzzy search function in the future?
Edit: Sorry I didn't see the first item in the TODO list
proud fable
oh someone is actually doing it since i first made my prototype lol
are you using whisper to transcribe the streams?
solemn grove
are you using whisper to transcribe the streams?
Yup, it's the best open source solution I could find but it still has a lot of flaws.
This is by no means an original idea haha, but is this something that exists already in the community?
proud fable
no, I made a prototype version a while ago, but haven't made the whole thing and transcribed everything
I used whisper turbo v3 and auto generated yt subtitles while available
solemn grove
I'm using large-v2 since it supposedly has the highest accuracy, admittedly even after I finish setting up an automatic pipeline it'll probably take months to transcribe the backlog of streams so I'll likely revisit this topic
proud fable
have you already started transcripting the backlog?
solemn grove
Very slowly yeah
proud fable
do you have a spreadsheet of the progress with all vods listed, I could help by donating some compute
also you could skip some vods where youtube auto generated captions are available
solemn grove
I've got ~15 handpicked streams in the db, and will upload the 2024 subathon (minus asleep segments) in a few days. Not in a hurry since this is a hobby project, but dumping the yt auto generated subs before replacing them w/ whisper transcriptions could be a nice thing to have, I'll look into it
proud socket
Yup, it's the best open source solution I could find but it still has a lot of f...
where are you getting all of the older vods? it was my understanding twitch limits vedal to 100
solemn grove
where are you getting all of the older vods? it was my understanding twitch limi...
The VODs are from YouTube, channels are linked in the footer of the website
solemn grove
The entire 2024 subathon (minus asleep segments) is now uploaded! Going to look into improving search next...
west plover
great work, this will be useful for wiki editors
solemn grove
Advanced search is now available! Though it's still pretty janky. Still figuring out what configuration makes the most sense
wintry herald
I put together a simple website for searching through past Neuro-sama stream tra...
Thank you! I could use that...
solemn grove
Added more search filters!
Archive is steadily growing, should have 2024-present complete by the end of next week
solemn grove
Archive now includes everything from the first subathon to now (will probably filter out the subathon asleep segments from the transcription at some point)
west plover
is there any plan to add a page 1 of x feature to the database to load more on demand, but otherwise save resources if not requested? or is that a budget limitation? I know when you search "the" for example only the most recent vod comes up so theres clearly a parse limit.
id also like to suggest a calendar date picker ui to autofill the date range in Search Options. its not necessary but its nice quality of life
whether or not my suggestions are feasible/something you want to implement, appreciate the work you've done 
solemn grove
whether or not my suggestions are feasible/something you want to implement, appr...
Yes both of your suggestions are something that I plan on adding!
west plover
is there any way to make searching numbers as the number and the word bring up the same result? like in your example on the home page searching “ten tin cans” brings up the miyune minecraft collab, but “10 tin cans” does not. 
solemn grove
is there any way to make searching numbers as the number and the word bring up t...
Ideally this becomes less of an issue when more features are added to advanced search, I'll toss this into the TODOs for now...
By the way results now support infinite scrolling
solemn grove
I'm excited to announce that the archive is now complete with 422 videos transcribed from the official and unofficial VOD channels!
Thought I'd share some interesting findings with the transcriptions:
I hope to keep the archive up to date as long as there are active users, lest it be... abandoned.
