This thread is for general userscript discussion, questions, recommendations, requests, questions, tips and tricks and sharing workflows, when not suitable for their own thread or one of the existing more-specific threads.
Maybe you should turn your initial post into a wiki post now, because after a while, you can no longer edit it.
But it will make it editable by everyone.
I didnāt find your thread when I searched - if you think this one duplicates it, I am happy to delete this one. While making this thread I found the topic for what I wanted to post anyway!
By the way, there is a bit of userscript and development conversation in the Discord scripting/coding channel (I know that doesnāt help people who donāt use Discord, and that it would be better in the open forums, Iām just sharing that the younger crowd seems to prefer to chat there)
Usually the releases I look at have all the links, or they donāt (in which case I add them all at once), so Iām just wondering if itās useful for me.
P.S. the recording-URL addition rate has jumped up again, I wonder if itās because of all the new scripts that can submit lots of relationships at once/submit and close tabs, etc?
Sometimes I feel making sure that the correct links were added. Track mismatches should be rare thanks to how Harmony harmonizes releases from different services but Iāve still ran on a few edge cases. Or there might be a prior mistake on the recrding chosen for a certain track within MB.
Other times I work with albums that have a handful of different releases (deluxe, expanded, clean/edited, etc.). I donāt think itās always worth adding all possible links to the clutter if itās the same (re-)master recording from releases distributed to the same area, so it helps me to check the prior additions.
Or, I might just wanna check a certain recording on Spotify or elsewhere and jump there directly.
Indeed, it could be mostly serving my niche needs for now, but I nevertheless felt like sharing it. For what itās worth, @chabanās feedback on Greasyfork taught me thatās itās possible to fetch all these links by a single API lookup on the release. You learn something every day.
[mod edit/move: this discussion relates to the release of the āMusicBrainz Quick Recording Matchā userscript, here]
Interesting. Iāve had cases where this would work great, and others where the first suggestion is very off. Maybe it would be useful to check the suggestion list for titles that are close to the wanted one and choose those, or highlight those tracks where the suggested recording artist / title differs significantly.
Do we have any? That script is scary as not everyone checks script output (see release dates on Digital Imports for examples).
There should be fines for mistakes. Build us up a slush fund.
Neat idea. Iād worry about a script as many people donāt check. There used to be a āguess worksā script before and it often made errors due to search being weird (try finding Roger Waterās āMoneyā and you get Abbaās āMoney Money Moneyā. Matching Dark Side of the Moon is always a challenge with tracks like Time and Money)
What would be good is a script that not only hit a match, but could then check that match to see if the recordingās artist has ever performed that work before. If not skip to the next until a match found.
I agree with @PacCeggowk9oc and @IvanDobsky ās concerns. Matching artist is also not enough as often thereās more than one recording of a song by the same artist.
If Iām adding a copy of a release I just select that as a duplicate first which brings all of its recordings and then apply the required changes.
Iāve often thought itās a pity that the inline search does not prioritise the context, like track artists, but I havenāt written a ticketā¦ Lazyā¦
(Sorry about the move of the posts, someone reported them and other people have also mentioned in the past that they would like that other thread just for new releases, which seems fine)
re. the concerns, the recording tab already displays all the checks that I would do - track/recording title, artist, and length.
However I went ahead and had Gemini add highlights for low confidence matches, some extra assistance for anyone who is the perfect combination of clever and idiot to be able to find and install this script, yet not look at the matches:
Nice idea , but would it not be better to just skip those and leave them unlinked? If a low confidence match managed to pick a track 33% longer and with a totally different name I think it is better to leave a gap in the matches instead of linking a mistake. If the script give 90% matches and the user has to fill in a couple of holes this still seems a big time saver. Better to have a new recording linked in there than something so far off.
Weāve all seen the editors who just charge through a submission and not worry about the quality of data.
(Please donāt take this as a negative bash. Iāve just been cleaning up a number of compilations where people are, for example, selecting a shorter 7ā instead of a 12ā with an obvious length differenceā¦)
I think that would be worse/give the illusion of accuracy- currently the script makes no pretense at being āset and forgetā. It automates a very simple task.
If a problem arises I can revisit, but I am willing to bet that it will be fine, if anyone else even uses it.
Edit: Though actually a toggle to auto skip low confidence matches could be useful for some people, if itās not too tricky. Watch this space (maybe!)
It would be too easy to take your money. Have you not seen what noobs with scripts are capable of.
Nice. I know you love talking to your AI pal We also need a Confidence Value for the editor tooā¦ āif not edited 1000 items, editor not allowed these elite toolsā.
I suppose thatās what is coming next - the AI Submission Bot who does this all for you. MusicBrAInz can read your artworkā¦
Edit: And a genuine feedback comment from looking at the code. Maybe make the confidence check based on a percentage error, and not a fixed number of seconds. Will then better adjust for longer recordings.
Edit2: being a bit of an evil testing person I tried to give your script a bit of a kicking. In a polite feedback way. Not highly scientific as I just picked two dozen Releases from my collection, unlinked everything, auto-linked everything. Even some nasty boxsets it got the lot.
Now this is a bit of a rubbish test as all these were already in the database. Be interested to see how it handles something new. Thatās the kinda time Iād expect it to say ānope, canāt find a matchā. Should have a box appearing in post anyday that will be good to test with. Will let you know
Thanks for testing! Seems like you found a quirk of this script - since it selects the first āsuggested recordingā, and if you āunsetā a recording it will go to the top of that listā¦ it will always re-select the previous recording, giving the wonderful veneer of perfection, when itās not really doing anything.
I donāt know if I would run this script on compilations, it was for my use-case of adding new releases into existing groups (if I donāt want to re-use the existing tracklist, so I can check/compare for minor differences).
But I had Gemini make some changes so that you can also try it on existing releases. You can now ignore/reset certain results, and choose between āfirst suggested recordingā and āfirst search resultā. I havenāt had time to test this much yet, but it seems to work fine:
I still feel that these options give the illusion of a āset and forgetā reliable tool, rather than just automating a simple series of clicks, but the people have spoken.
The confidence thresholds are indeed extremely clumsy but I donāt have the headspace to improve them right now, since it needs a bit of thought put into it. Contributions welcome.
I hate it so muchā¦ I sacrifice my time and nerves in the name of data!
I would never use AI to do something that someone actually wants to do. But itās welcome to do some shitty code to make data entry easier. It took at least hour of smashing my dumb head into Gemini to have it get these updates right.
This is part of the problem. Other users are never you. And they gonna use it on bad cases.
Yeah, I is awkward like that. Didnāt surprise me that I found a quirk as it seemed too perfect. I didnāt read all the code, but it did feel like something was cheating somewhere.
Will add feedback from new releases I add. See how it behaves. The more we break it, the better it can get.
Could this script be extended to also work on the āEdit relationshipsā page when adding existing works to recordings ā¦. most of the time for my types of edits its the first suggested work Iām interested in.
That script is scary as not everyone checks script output (see release dates on Digital Imports for examples).
Build us up a slush fund.
, but would it not be better to just skip those and leave them unlinked? If a low confidence match managed to pick a track 33% longer and with a totally different name I think it is better to leave a gap in the matches instead of linking a mistake. If the script give 90% matches and the user has to fill in a couple of holes this still seems a big time saver. Better to have a new recording linked in there than something so far off.
We also need a Confidence Value for the editor tooā¦ āif not edited 1000 items, editor not allowed these elite toolsā.
can read your artworkā¦