chromaprint VS youtube-dl

you could always cut up the video into segments A,B,C,D,E using like ffmpeg -ss 00:00:00 -t 00:10:00 -i input -c copy A.ext (where A is the first 10 minutes of video) and then recombine with https://trac.ffmpeg.org/wiki/Concatenate, might take a while though and some basic math. maybe you could do something programatically with chromaprint by identifying the repeating audio segments? that's just an idea since that's what this jellyfin plugin does to identify repeated segments of shows in the form of opening credits.

I'm currently doing the same. I'm using the chromaprint library via the pyacoustid library, using the technique described here, which should give a certain amount of leeway in terms of pitch and other variables. See also the notes in the chromaprint documentation about truncating the samples to reduce noise.

There is a music fingerprinting project called Chromaprint. The fingerprints generated by this library are widely used in music database, including Musicbrainz.

For me, the less-complicated client called "DVD" is more reliable for downloading than the client called "Newpipe"; the later stops working intermittently. YMMV.

(Neither are as reliable as using own custom command line programs on desktop, or mobile via Termux. If Google makes a change I can fix/workaround immediately. No waiting for app developers.)

For me, Newpipe works well as a Soundlcoud client.

From the Kotlin source for "dvd":

Video downloader app powered by yt-dlp (formerly youtube-dl).

Features

Download video/audio from 1000+ sites supported by yt-dlp.

Run custom yt-dlp commands.

Update yt-dlp version from settings.

Share link via other apps.

References

dvd - https://github.com/yausername/dvd

youtubedl-android - https://github.com/yausername/youtubedl-android

yt-dlp - https://github.com/yt-dlp/yt-dlp (formerly youtube-dl - https://github.com/ytdl-org/youtube-dl)

Yep. yt-dlp and youtube-dl

https://github.com/yt-dlp/yt-dlp

https://github.com/ytdl-org/youtube-dl

Will also start to feel the impact. My theory is that we will see a bunch of new video hosting sites as youtube itself attempts to lock down its ecosystem. They haven't paid attention in any adversarial way as far as I can tell.

When they do, it wont be great.

Youtube-dl still functions quite well. You can even grab the CLI here: https://github.com/ytdl-org/youtube-dl

And occasionally you can find one of their TLDs just by doing a search on DDG or Brave.

If you don't want to switch to yt-dlp, then wait for a fix in the code of youtube-dl and then download a version from Unofficial daily builds.

I'm using fingerprints built from different versions of the song, including the original and remastered versions you can find on the Discord and the "pitch corrected" version you can grab from YouTube. I'm chopping each fingerprint into 100 sample chunks and searching each YouTube video (audio only, using youtube-dl) for them. Checking for the fingerprint is the fastest part of the process, so trying many against each download isn't a problem. I've confirmed it's possible to detect TMS in videos about TMS (eg. Whang, CW Schultz) where less than the full song is played.

It's getting the point where I wont use YT unless I'm using https://github.com/ytdl-org/youtube-dl

I downloaded the video from YouTube using youtube-dl.