I’ve been using Music Brainz Picard for a few years now and back then I had a ton of unorganised audio files. Now I have, what I like to think of as, a pretty sustainable system for collecting / processsing (i.e. listening) / and organising my music
I think I calculated it was going to take the best part of 15 years to organise all my music, but nevertheless it’s piece of mind the system is attune with my consumption behaviour and I’m not paving cow paths again
Based on my knowledge of the software, I’ve created and refined this mindful, non-destructive (i.e. best for keeping releases intact), workflow and wondered if anyone uses anything similar; or had any feedback?:
2. Process http://musicbrainz.org/ ~ https://picard.musicbrainz.org/docs/guide/
2.1 Gold pass
Cluster
Lookup
Scan
Save
Lookup
Scan
Save
Scan
Save
2.2 Silver pass
Cluster
Lookup
Scan - remove/sanitise duplicates ~ https://picard.musicbrainz.org/docs/basics/
Save - merge to gold/trash
Lookup
Scan - remove/sanitise duplicates ~ https://picard.musicbrainz.org/docs/basics/
Save - merge to gold/trash
Scan - remove/sanitise duplicates ~ https://picard.musicbrainz.org/docs/basics/
Save - merge to gold/trash
2.3 Unmatched pass
Add/create release http://musicbrainz.org/release/add ~ http://discogs.com/release/add
Feel free to dismiss the numbering, since this section is part of a larger workflow
Thanks for sharing. I’m having a little difficulty understanding your workflow.
So, for your gold pass, load some unmatched files, cluster, then lookup the cluster or remaining unmatched files? Then what are you scanning, because I can’t scan files once they’re matched to albums? If I get this, I might begin to understand the rest of it
Ultimately, “Scan” exists for completeness, but it’s useful on occasions to ID stray files (i.e. files not contained in folders, which reside in “Unmatched Files”)
Sometimes I find Music Brainz Picard will ID a completely misnamed track correctly
When this occurs, with a bit of detective work (e.g. referencing the enclosing folder/s context), I will usually discover the track is part of a bootleg release. This usually, then, gets processed during a later pass (e.g. “Silver/Unmatched pass”).
Sorry, but I still don’t understand your process because I don’t know what things you’re doing the different actions to. I’m also not sure what you mean by ‘gold pass’ ‘silver pass’ etc. (I guess they refer to the colours the CDs go in Picard depending on the strength of matching?).
So for example, load one folder, cluster some unmatched files, lookup the remaining unmatched files, scan the still remaining unmatched files, save any gold releases? Then lookup the clusters, scan the clusters if that didn’t work? etc. etc.
I also don’t understand your link for remove/sanitise duplicates because those docs don’t mention duplicates…
Sorry
Personally, I like to scan everything and submit once I’m sure it’s matched so others can use acoustids in the future; so my routine would start with scanning everything and then pulling stuff back out of the release side if it had mis-matched.
I use the term “pass” with reference to shaving - sorry for the ambiguity; I thought it would be a recognised term in this context
“Since cuts are more likely when using safety razors and straight razors, wet shaving is generally done in more than one pass with the blade. The goal is to reduce the amount of hair with each pass, instead of trying to eliminate all of it in a single pass. This also reduces the risks of cuts, soreness, and ingrown hairs. Alum blocks and styptic pencils are used to close cuts resulting from the shave.” - https://en.wikipedia.org/wiki/Shaving
The goal of a “Gold pass” is to save out gold ID releases, and the same with “Silver pass”. “Unmatched pass” is generally where the automation stops and one has to start adding/creating releases to the DB - it’s far more considered here.
–
Sometimes a silver ID release will be silver because it contains duplicate tracks - you will need to prune these to save out a gold release. This is actioned during a “Silver pass”.
–
Referencing my original post (i.e. “non-destructive)”: scanning is [potentially] the most destructive feature you can run on your files; especially if it’s used on a first pass and without clustering.
I think I get what you’re doing and each to their own, but I prefer to manually complete releases, rather than undo them - in theory it’s less work and you’re less likely to have to deal with automation errors
If you give the logical passes a shot you may find that flow works for you too; with the addition of many degrees of safety to [batch] workflow
Thanks for the explanation of passes. It’s what I would have guessed.
I tend to work on a single album at a time and it sounds like you’re working on dozens so I understand the more cautious approach. I still think you could start with a scan and select all that’s matched and drag it back, but perhaps the way scan is integrated could be better achieved.
Without wishing to sound like a broken record I still don’t understand your process so I can’t yet give it a go. After loading a dozen albums and pressing cluster on the unmatched files, I don’t know what I’m pressing lookup on, etc.
… my screencast time ran out, but after “Lookup” I would run a “Scan” before saving out any gold releases
The next step would repeat the process, but this time without “Cluster”:
Lookup
Scan
Save
The final step would be without “Lookup”, just “Scan”, to complete a gold pass
Scan
Save
In a similar manner, silver and unmatched passes would come next - on these occasions you’ll probably need to put in some manual labour to achieve gold releases
Your passes make sense to me. I would recommend that you setup a file naming scheme so that the gold releases could be saved in separate folders. That will leave you with only the ones that are still silver to work through. It helps to see what’s left to do.
Also, if you are pulling cover art, you’ll definitely want each release in a separate folder.
If you start with a “scan only” pass, “expand all” and look at the “red” (changed) files first, you have a kind of an extra “sanity check”. The way you do it now, that possibility gets (mostly) lost. Even if you drag everything back and do it the way you’re used to from there, it can still be useful as a pre-check to catch some problems.
Load dozens of albums
Cluster unmatched files
Lookup clusters
Scan remaining clusters
Save gold discs
<- I usually restart MBP here
Lookup unmatched files
Scan unmatched files
Save gold discs
Gold releases have always been pretty accurate for me TBH
TBH this isn’t something I’ve been doing, so, yes, I’ll be more mindful of red releases in future before I start my method - cheers
However, when I initially add files to the left pane MBP automatically generates red releases in the right pane. After running a scan I get no additional red releases, so running the scan first doesn’t appear to offer much…?
How often do you find MBP discovers additional red releases (I’m guessing often enough to make it part of your process)?
Those probably already have MBID’s attached, so they automatically get “shoved” right if you have that option enabled. If you want to “sanity check” your whole collection just to give it a try, you could temporarily set the option “ignore MBID tags when adding new files” and run a scan on it, to see what would change (just make a backup or be very careful not to safe anything ;)).
To clarify: I don’t mean red release icons - those are just error messages (like p.E.: “couldn’t load album art from X”). I meant the red tinted recordings: The more tags would get changed, the stronger the red tint of that recording is. This is why I use acoustid / scan initially. Most of the slightly reddish tinted recordings will just be updated data (year, accents added, featured artist updated etc), but then there’s some that will be deep red, because the tags from the fingerprint match are totally different from how it was tagged before:
(problems with the files: )
File has (initially) been tagged VERY incorrectly
Tracks have been switched around somehow
Recording is actually from different artist / cover version
File is broken / contains silence from a ghost track and bonus track that shouldn’t be there for the release it was tagged as
or (Problems with musicbrainz DB):
Someone added an acoustid to the wrong recording
The acoustid has simply been added to the same recording from a different release group (p.E. compilation), in which case the information attached to your files was right but sometimes it might make sense to merge recordings online if noone already has.
How often you get those depends mainly on where you get your music from. If it’s a competent & trustworthy source, the files have been tagged well before and you focus on complete / whole albums from widely known artists, you might not discover anything useful except that “those files are probably really in good shape!”.
If on the other hand you get “mixtapes” from random people sometimes or something like that, this could help detect a lot of mis-identified recordings.
Either way - it doesn’t usually take long since I can just “expand all” + scroll through it, then drag it all back if there’s no problem. I’d guess when I first add files, for every ~300 files from ~100 different artists, I find about 1 “terrible mixup” (totally wrong release + band + title) that way that I wouldn’t have had a chance to catch - ever - otherwise… and that alone makes it worth that extra step in my case.
Just scanned some of my existing files again just to see if the acoustids that have been submitted since I last checked changed anything. In ~2000 files from ~1000 releases that have already been checked several times before, I found:
7 recordings that were tagged as the correct work but wrong recording (/ mostly different intros/outros, radio version etc : same release group but release from a different country and something was bleeped etc)…
1 that was a cover / “reinterpretation” from an entirely different (less known) artist than what the tags said…
2 where my tags were correct but someone else seems to have submitted the acoustid from a wrong track (have yet to figure out how to report that)…
1 that was recognized as an obscure live bootleg by “lookup” somehow when really it was just the (non-live) “album version” from a compilation (no idea how that slipped by me - had that file for years)…
1 file where the last few seconds were simply missing and I thought it was supposed to end that abruptly :D…
several recordings that should be dupes in the MB DB (have yet to figure out which ones should be merged)
NB I’ve included some stats in the top right [search] form field
Don’t think I’ve experienced much red activity, which isn’t “Could not load album…”, but I will be more vigilant in future and perhaps follow up this thread with any questions if that’s okay?
In any case, I do certainly get better, initial, results from using [checking] “Automatically scan all new files” and “Ignore MBID when loading new files”, so thank you for elaborating on @tommycrock’s original point
Gold and Silver are just the coloring of the discs in Picard. If Gold, it means there is exactly one match for every track (possibly with some special tracks excluded, like data). Silver means that you don’t have track matched perfectly - could be missing some tracks, missing an entire disc of a set, or possibly duplicates for a single track. @ldexterldesign is using that matching to describe his process of gold and silver passes.
Thanks