Your way for handling non-ASCII artists and tracks names?

[tukoz]

beets 1.3.16 on Arch linux here. Casual user slowly getting in, consolidating a music collection with ~1K Oriental pieces (mostly Arabic, Chinese and Japanese). Beets works rather fine for Western artists :) But importing the non-Western tracks hardly works, probably because of a wrong set-up or methodology on my part.

• For non-Western Artists and Tracks with some reliable metadata, I barely get over (50%) but As Tracks often works. An issue is that beets takes the non-ASCII names, which I don't want (theoricaly fine but hinders further use on any system/device with a latin keyboard or poor unicode support). This keeps me from writing any ID tag or filename into the files --and so, to clean the "Latin" part of them. E.g:

David Bowie Ryuichi Sakamoto - Merry Christmas Mr. Lawrence (Furyo).mp3 (1 items)
Correcting tags from:
    David Bowie Ryuichi Sakamoto - 
To:
    坂本龍一 - Three         # unreadable (conjsole)/searchable
URL:
    <snip>.
(Similarity: 12.9%) (missing tracks, album, artist, tracks) (CD, 2013, Decca Records)
 * Merry Christmas Mr. Lawrence (Furio) (#0) (5:46) -> Merry Christmas Mr. Lawrence (#10) (5:35) (title, length)

• For those with broken metada, detection does not work even with tracks (unknown album and else lent by friends while travelling). And Accoustic fingerprinting is often errr, funny:

    $ beet import -alpW POP/VA\ Best\ Arabic\ Music\ Collection\ Of\ All\ Times/
    ...As Tracks
    POP/VA Best Arabic Music Collection Of All Times/SHATER..REMIX.mp3
    Tagging track: DJ Smash - Space Dust
    URL:
        http://musicbrainz.org/recording/dbe4628a-817e-4af8-8fe7-41e91e73c58d
    (Similarity: 100.0%)

    POP/VA Best Arabic Music Collection Of All Times/بنوتة.mp3
    Tagging track: Frank Sinatra - Nancy
    URL:
        http://musicbrainz.org/recording/419894fb-6dd6-4897-a269-7a09fcb45b67
    (Similarity: 100.0%)

So a Palestinian dabke (rythmed dance) is now known as Franck Sinatra's daughter name? hmmmmm pffffff :'-)

Is there a way to make Beets first deal with the best matches? Or shoulld we manually guess-and-move in a different location, eg. all files that could take profit of an Acoustic fingerprint or manual search?

I'd like to hear of your set-up and or strategy for handling non-ASCII art. E.g. in term of file structure before importing, config.yaml/overrides and plugins.

[sampsyo]

An issue is that beets takes the non-ASCII names, which I don't want (theoricaly fine but hinders further use on any system/device with a latin keyboard or poor unicode support).

If you like ASCII-only filenames, you might like the asciify_paths config option.

For those with broken metada, detection does not work even with tracks (unknown album and else lent by friends while travelling). And Accoustic fingerprinting is often errr, funny:

Yes, acoustic fingerprinting can have false positives.

Is there a way to make Beets first deal with the best matches? Or shoulld we manually guess-and-move in a different location, eg. all files that could take profit of an Acoustic fingerprint or manual search?

I don't think I quite understand—do you want to deal with high-confidence matches first and then low-confidence matches in a separate pass? If so, then yes: configuring beets to move files after it imports them seems like a good strategy. Just make sure you have a backup in case it does something unexpected!

[tukoz]

If you like ASCII-only filenames, you might like the asciify_paths config option.

Option does a great job, along with language for preferred aliases. Clearly I overlooked them before. An option to get the artists/art's names in both original and Latin languages would be optimal (like how IMDB/the TVDB do for movies and documentaries), but I bet it's not feasable with ID Tags?

Is there a way to make Beets first deal with the best matches?

I don't think I quite understand—do you want to deal with high-confidence matches first

Yes as until now I've been stuck at every two tracks and at 2 or 3:00 AM found myself lacking the time to import over a fraction of my Music library. Now, none_rec_action: skip also helps the import workflow. This way I can import best matches first in the collection, and deal with the more « obscure » (from a Western PoV) later on (hundred of Skip/Enter search/as Tracks avoided).

Conf file's much longer now, with import time decreased by a level of magnitude!

[sampsyo]

Yes, there's not currently a good way to keep multiple languages for the same music.

I'm glad things are working out OK!