forum

[Proposal] Further Metadata Amendments/Fixes

posted
Total Posts
37
Topic Starter
Noffy
Hello from the joined efforts of Lanturn, Oko, Doormat, Noffy, we bring a new m e t a d a t a

Update: Read the draft that resulted from this thread over at github



old
Metadata thread #999,
I'm sick of seeing them too, yeah

Thanks celerih for being my discussion buddy and keeping me sane while doing this, or else this post would be even more visibly annoyed than it currently is.

this is too much to fix without a thread, please give feedback to make sure any suggestions in this thread would make things better instead of worse. If they'd make it worse, give suggestions to fix that. Thanks.
Or complain about other rules that suck, if you want. Suggest fixes to those too, if you want. Let's fix those instead of complaining about it without doing anything if it's really so blatantly bad.


Delete Circles

For songs belonging to doujin circles, the circle name must be used over the vocalist or composer, unless these contributors are not part of the circle. In these cases the priority falls on vocalist followed by composer for instrumental songs.
Instead of making things simpler, this has made the artist field more complex due to the research required to determine whether those who made a song are actually proper members of the circle or just guests for that one release.

there's 1 or 2 things we could do

1.) delete circle rule entirely (leaving it up choice based off what is more sensible to the specific song or consistent with previous maps.. like it was before)
2.) make any release under a circle be credited to a circle regardless of whether the song's creators are proper members of the circle.


This is weird

If a song or artist are referred to in multiple ways on official sources provided by the artist, the mapper is free to choose any of the romanisations. The only exception to this is if the song already has a mapset in the Ranked Section, in which case the corresponding guideline applies to it.
I don't think this was supposed to say romanisations, just apply to metadata in general, oops...
honestly, is this really needed? it's not like we say you must use a specific kind of source to reference for metadata to begin with.

There's a loophole here
Any form of CV: such as CV,, Cv, Cv. and the likes are to be written as CV: only. If only the character is credited in any sources, the CV: marker has to be added to satisfy traceability to existing people. If such a marker is added to an artist, they must be added following the fictional character the character voice actor represents.
This accidentally lead to a loophole that Character (Voice Actor) does not require CV: to be added.
Should probably fix that to just be Character (CV: Voice Actor) in all cases.

suggestion:

Whenever a fictional character or characters are credited as the singer of a song, the artist field is to be formatted in a Character (CV: Voice Actor) format, where "Character" would be the character name and "Voice Actor" the name of the actual person behind them.
Source is a headache

You must use the Source field if the song comes from, is remixed from or specifically fan-made for a video game, movie, or series. Website names are not an acceptable Artist nor Source.
When the song has appeared in multiple medias, even after its initial release, it may use the source that the mapset is themed around (Backgrounds, Storyboards, Videos, etc.) as long as the song itself appeared in it.
If a series applicable to be used as a source is clustered into multiple sub-series, the most precise label should be prioritised.
we somehow ended up splitting source into 1 rule and 2 guidelines that half do the same thing as the old rule and half make it more of a headache, because sometimes it's really hard to determine if a song came from the source or if it was put in there later (making the source only usable if the map is explicitly "themed" around it)

suggestion
delete these and revert to the old rule, thanks

This needs revision

Use the metadata of a song as is without adding any additional markers, such as Full Version and the likes. If your track is a mashup or something among those lines, use whatever the artist supplies while paying respect to proper English syntax rules.
It directly contradicts the TV Size rule demanding you add TV Size

It also directly contradicts
If the creator of the mapset has done major edits to the .mp3, they are free to name it appropriately to signal that this song is a special version. In this case the original songs must still be clearly indicated in order for players to be able to search for the original songs.

which says you CAN add stuff to it, as long as you edit it enough.

suggestion to fix this
Change the rule "Use the metadata of a song as is..." to
Use the metadata of a song as is without modification, unless the modification is specified or allowed by other rules in the Ranking Criteria
Robots
celerih has pointed out that
The artists of a song must be traceable to existing people. As such, programs like Vocaloids or fictional characters cannot be used as the sole artist of any given song.
does not account for songs that may be entirely composed by robots/ai. I think it would be sensible to allow crediting of the robots/ai in such cases, but I personally can not figure out a good wording to account for that.
Any ideas?


Delete this
Special characters that have both a full-width and half-width equivalent must retain their original printing forms if they are used in the unicode artist fields, so that searching while copying and pasting the original song title is always possible. This does not apply to additional information that is printed in half-width to begin with because searching for these will be ambiguous no matter what.
a.) the rule that the metadata shouldn't be modified unless it is done to follow the rules would already cover this
b.) the rules in metadata about standardizing cv tags, tv size, vs, etc, etc, etc, already make copy paste searching more difficult. So it doesn't really help in that aspect.
celerih
Went through these already with noffy and I fully support these changes. They help simplify very convoluted rules and would help unify a lot of scattered rules in the current metadata RC

Edit: gonna be adding some stuff about the (TV Size) rule soon, since I believe it needs to go too
anna apple
I support you guys !!
Stefan
rip SiivaGunner
Cynplytholowazy
:) Supportive to these changes
Uta
agreed with these aswell
Xinnoh

Noffy wrote:

suggestion:

Whenever a fictional character or characters are credited as the singer of a song, the artist field is to be formatted in a Character (CV: Voice Actor) format, where "Character" would be the character name and "Voice Actor" the name of the actual person behind them.
Under that it would mean Robbie Rotten or Troy Bolten would require the CV thing since they are fictional which doesn't make any sense. This should only apply to animated stuff with voice actors.

"But why not just do it for anything with voice actors?"
High School Musical 1 didn't actually have the actors sing their own songs, they were voiced over so it would technically require CV if we did it based solely on voice acting. (why do i know this useless trivia)

otherwise good
7ambda
make metadata great again
Topic Starter
Noffy
Thanks for the support guys, but don't forget to give it a critical look to make sure all of the suggestions would be doing what they intended, and not have any unintended side-effects.

Really appreciate it \o/

Sinnoh wrote:

Noffy wrote:

suggestion:

Whenever a fictional character or characters are credited as the singer of a song, the artist field is to be formatted in a Character (CV: Voice Actor) format, where "Character" would be the character name and "Voice Actor" the name of the actual person behind them.
Under that it would mean Robbie Rotten or Troy Bolten would require the CV thing since they are fictional which doesn't make any sense. This should only apply to animated stuff with voice actors.

"But why not just do it for anything with voice actors?"
High School Musical 1 didn't actually have the actors sing their own songs, they were voiced over so it would technically require CV if we did it based solely on voice acting. (why do i know this useless trivia)

otherwise good
I'm not entirely sure I understand the issue with this, as that would still be properly giving credit where it is due when the singing is credited to just the character.
In We Are Number One's case, it's actually credited to Stefan Karl Stefansson (the person) so this wouldn't apply. Other cases where it's credited to the voice actor/actual person directly shouldn't be impacted by this either.

If that is not clear, I'm not sure of a better wording to make it clear. Does anybody have a suggestion?
celerih
Okay so here are a few propositions from me. First of all

Romanise artist names in whichever order they are printed in the unicode fields. The sole exception to this are artists widely known and using a differing English alias.
This rule has been getting attention since it's essentially asking to often ignore official preferred romanization of artist's names. Especially for someone with little knowledge on japanese, knowing which order to put things in can be really annoying and confusing. In addition, most artists have a preferred romanization on their website which would make dealing with these artists a lot easier for most people. Now, one of the issues this had in the past is that extremely obscure things have been used as "official sources" which made things especially hard.

So here's what I propose: Change the rule to
Romanise artist names according to preferred romanization found on official sources. If none is provided, romanise artist names in whichever order they are printed in the unicode fields
In addition we can add a Glossary term for Official Sources: Official Websites, CD Scans, Album listings with labels, or official uploads to websites like bandcamp, youtube, or soundcloud. Third party websites for sales of songs would not be used as references, such as Spotify, iTunes, or Amazon.

This would help immensely what can be used as official sources and limit what needs to be searched when trying to find metadata. As well, we often see people unsure as to what counts as an appropriate source to use for metadata, so this would help make it clearer to them
Zhuriel
having the circle rule at all doesn't make much sense because circles can function both like bands and like labels. for example, the current wording of the rule would require many camellia songs to be credited to "KamelCamellia", noah songs to "ExistRuth" and laur songs to "Last Labyrinth" which does not serve any purpose except to confuse. circle names should only be used where circles function more as bands and the songs (or at least albums) are officially credited to the circle.

removing the rule entirely would make the most sense imo since coming up with a wording to cover all the edge cases the japanese manage to come up with would probably be a massive undertaking
UndeadCapulet
circle rule is currently better as is than it would be with either suggestion. simplicity is not better when that simplicity is taking out actually sourcing the creators. just sourcing the singer is not sourcing anyone that actually made the song, it'd be like sourcing the lead singer of a band, which is why the circle rule was implemented to begin with. and sourcing just the circle is now no longer sourcing the singer, which also doesn't make much sense, because they're the singer. it really doesn't take that long to google if the singer is a part of the circle or not.

if it really has to change i guess go with the circle, but the current rule is much better imo s:

other stuff sounds good tho i guess
celerih
Main issue with the circle rule is that now actually more often than not circles don't get credited, as very very very very often (Lasse can attest to this) vocalists aren't part of the circle, mostly producers and composers are. So you get this weird situation where you have a rule that's supposed to credit circles more but as a result circles are used less often as artist.

One example is my map here https://osu.ppy.sh/s/777789
While Hotaru is a very present in Halozy's song, she's not officially part of the circle despite having worked with them countless times (see: http://halozy.com/profile/ ) This would force me to use her as artist over Halozy.

SO yeah while in theory it's nice, in practice it really hasn't had the desired effect, which is why the rule is being changed
Nao Tomori
why not just make it circle feat. singer each time the vocalist isnt in the circle lmao that way both get credit
celerih

Nao Tomori wrote:

why not just make it circle feat. singer each time the vocalist isnt in the circle lmao that way both get credit
actually like... yeah. It's worked well with RD-Sounds in the last few months, this would help deal with the mess that is circles. Although you have the potential issue of stuff like https://osu.ppy.sh/s/790310 where you have a composer that's well known plus a circle plus a vocalist

For now we can use this rule as a base and improve upon it:
Guideline:
When a song is released by a circle, the circle name is preferred for the artist field. If a composer is independently well known outside of the circle, they can be used as artist. In the case of songs with vocalists, the following formats can be used: [Circle] feat. [Vocalist] or [Composer] feat. [Vocalist].
Lasse
the main issue with the current circle rule is that in so many cases vocalists and/or composers are not part of the circle, I looked at 5 random qualified/bubbled touhou maps and all of them would have to use vocalist only as their artist with the current rule, so you end up just ignoring circle, composer, etc.
basically ends up being like the old rule, but forcing vocalist as artist instead of making it an option in a lot of cases

for reference:
https://osu.ppy.sh/beatmapsets/798509/discussion/-/generalAll#/375287
https://osu.ppy.sh/beatmapsets/782106/discussion/-/generalAll#/375290
https://osu.ppy.sh/beatmapsets/790310/discussion/-/generalAll#/374527
https://osu.ppy.sh/beatmapsets/751553/discussion/-/generalAll#/375244
https://osu.ppy.sh/beatmapsets/785837/discussion/-/generalAll#/375204
Asaiga
Hi,

What about covers? For example DECO*27 feat. Yuikonnu or namirin*. Should we use circle name feat. singer in the artist field or just solely artist that does the cover? Sorry if I have missed the point somewhere.
Xinnoh
for the fictional character thing, CV is almost exclusively used to describe voice actors. Including actors with this doesnt make sense, best to just list the actor rather than the character
UndeadCapulet
ah i misread what the current circle rule actually did, my b
yeah i agree w/ [circle] feat [singer] being the best solution
neonat
That still seems weird when it involves a few or individual singers who make albums with multiple composers for their albums, using just the vocalist alone in the artist field is better tbh

Examples I can think of are most known utaites and doujin vocalists. While at it, many people like to take songs off youtube and all where stuff like this is uploaded, are you going to apply this to them as well? You cannot deny many of them have their albums and also sell things under a circle name to represent their booth.
Okoayu

Noffy wrote:

Delete Circles

For songs belonging to doujin circles, the circle name must be used over the vocalist or composer, unless these contributors are not part of the circle. In these cases the priority falls on vocalist followed by composer for instrumental songs.


Instead of making things simpler, this has made the artist field more complex due to the research required to determine whether those who made a song are actually proper members of the circle or just guests for that one release.

there's 1 or 2 things we could do

1.) delete circle rule entirely (leaving it up choice based off what is more sensible to the specific song or consistent with previous maps.. like it was before)
2.) make any release under a circle be credited to a circle regardless of whether the song's creators are proper members of the circle.

Fuck that rule imo. Leaving it up to choice is meh, leaving it up to "most common way" to do it is meh too because it increases resarch effort

sayhing vocalist > composer unless the circle is known by its own name a lot (ex. iojjj - Torikago vs Esper Gakuen - Torikago: no one in hell will know either of those unless they played my first rnaked map but iojjj is more well known in this community because i have two maps using her instead of the Circle) vs lets say ALiCE'S EMOTiON or C-CLAYS or whatever who are well known on their own

if we go for searchability i think reverting is best so this was a long yeah fuck i actually agree

Noffy wrote:


This is weird

If a song or artist are referred to in multiple ways on official sources provided by the artist, the mapper is free to choose any of the romanisations. The only exception to this is if the song already has a mapset in the Ranked Section, in which case the corresponding guideline applies to it.


I don't think this was supposed to say romanisations, just apply to metadata in general, oops...
honestly, is this really needed? it's not like we say you must use a specific kind of source to reference for metadata to begin with.

Nah, just romanisations we did that intentionally. This doesn't add much for the actual doing, but it clarifies the most common question you get when checking metadata because it says the first officially listed thing you find is usable instead of leaving this up in the air

if you still think this doesnt have its place explain further

Noffy wrote:

There's a loophole here
Any form of CV: such as CV,, Cv, Cv. and the likes are to be written as CV: only. If only the character is credited in any sources, the CV: marker has to be added to satisfy traceability to existing people. If such a marker is added to an artist, they must be added following the fictional character the character voice actor represents.


This accidentally lead to a loophole that Character (Voice Actor) does not require CV: to be added.
Should probably fix that to just be Character (CV: Voice Actor) in all cases.

suggestion:

Whenever a fictional character or characters are credited as the singer of a song, the artist field is to be formatted in a Character (CV: Voice Actor) format, where "Character" would be the character name and "Voice Actor" the name of the actual person behind them.

yeah, oops wow let's go with that suggestion
ppl totally got the intent behind it and still fiddled around? whew it's almost like you need to spell everything out and then nobody gets it anymore

Noffy wrote:

Source is a headache

You must use the Source field if the song comes from, is remixed from or specifically fan-made for a video game, movie, or series. Website names are not an acceptable Artist nor Source.

When the song has appeared in multiple medias, even after its initial release, it may use the source that the mapset is themed around (Backgrounds, Storyboards, Videos, etc.) as long as the song itself appeared in it.

If a series applicable to be used as a source is clustered into multiple sub-series, the most precise label should be prioritised.


we somehow ended up splitting source into 1 rule and 2 guidelines that half do the same thing as the old rule and half make it more of a headache, because sometimes it's really hard to determine if a song came from the source or if it was put in there later (making the source only usable if the map is explicitly "themed" around it)

suggestion
delete these and revert to the old rule, thanks


This was never intended as such, at least I meant it as "as long as it's used in X it's fine" as long as it's obvious that whoever used it had to actually license it

Noffy wrote:

This needs revision

Use the metadata of a song as is without adding any additional markers, such as Full Version and the likes. If your track is a mashup or something among those lines, use whatever the artist supplies while paying respect to proper English syntax rules.


It directly contradicts the TV Size rule demanding you add TV Size

It also directly contradicts
If the creator of the mapset has done major edits to the .mp3, they are free to name it appropriately to signal that this song is a special version. In this case the original songs must still be clearly indicated in order for players to be able to search for the original songs.

which says you CAN add stuff to it, as long as you edit it enough.

suggestion to fix this
Change the rule "Use the metadata of a song as is..." to
Use the metadata of a song as is without modification, unless the modification is specified or allowed by other rules in the Ranking Criteria


Depends on the order you read it in, honestly - always specifying that TV Sizes are exempt seemed like so much mroe text because i didnt know how many rules will end up ignoring this lul

Noffy wrote:

Robots
celerih has pointed out that
The artists of a song must be traceable to existing people. As such, programs like Vocaloids or fictional characters cannot be used as the sole artist of any given song.

does not account for songs that may be entirely composed by robots/ai. I think it would be sensible to allow crediting of the robots/ai in such cases, but I personally can not figure out a good wording to account for that.
Any ideas?


the one circle of people or person making the AI that made the music created the music, easy


Noffy wrote:

Delete this
Special characters that have both a full-width and half-width equivalent must retain their original printing forms if they are used in the unicode artist fields, so that searching while copying and pasting the original song title is always possible. This does not apply to additional information that is printed in half-width to begin with because searching for these will be ambiguous no matter what.

a.) the rule that the metadata shouldn't be modified unless it is done to follow the rules would already cover this
b.) the rules in metadata about standardizing cv tags, tv size, vs, etc, etc, etc, already make copy paste searching more difficult. So it doesn't really help in that aspect.


Suggestion: reword it into the requirement to tag as many similar or different ways to write your shit as possible if you find double width shit
Topic Starter
Noffy
The intro post has been updated with a link to a github draft! (thanks Oko)

It may not fully reflect discussion here yet, but that's because we should preferably have more discussion first.

Due to the large number of changes, the draft displays the entire metadata sections as it will be in the RC, (except for the angry red triangles)

Let's work this out together :)
Nevo
I still don't understand why (TV Size) must replace an existing marker in a song it kinda contradicts the entirety of the rules. You are practically saying well we want to follow the metadata of a song from what the artist uses, yet you are changing something the artist picked because it's inconsistent with what other artist do.

Tldr
Tv size rule is dumb and is inconsistent with the entire reason for these rules.
Lanturn
The whole point of the prefered artist order and TV Size marker is to aim for a more standardized approach and it was one of the solutions I came up with. Something that would be fair. Other options for the artist order were to use First Name Last Name, or reverse for everything. Or go back to the Eastern Artists must use Family Given thingy we had in the past.

Standardized metadata is being adapted to make it easier to apply metadata. Take a look at wikipedia. They always list singers in a First Name, Last Name basis, regardless of country of origin. Other database / fansites will always have their own preference as well. So instead of having to look for obscure website reference of what specific TV Size or Artist order preference, we attempted to build our own rules to make metadata more streamlined and simple with a standardized process.

I'll say one thing. It's a hell of a lot easier to google search the Japanese letters of a singer and look at the results. http://puu.sh/AKn1N.jpg for example, and just simply reverse the order, rather than have to dig deep into a ton of different websites and maybe find a preferred romanization order. This is the whole point of standardized metadata.

"Delete Circles"
Circle feat. singer/composer definitely seems like the best standardization method at the moment. I'm down for this.

"This is weird"
@Oko. This could also apply if some unicode uses different characters or even capitalization options. It should definitely say metadata in general. Bombs away for example does have a release as Bombs Away on the back of an official CD cover so the user could choose either of these to use in the unicode field.

"There's a loophole here"
As discussed, yes this needs to be added.

"Source is a headache"
We could probably remove the whole theming part and just let the mapper choose any official source. But I'd still like to see a themed map use that specific source rather than something else. Say you have one themed around Hatsune Miku's Sega games but then use a SVDX source. Right, it doesn't really make sense.
Let's say we revert back to the old way. Do we want to enforce the most precise source or stick with the optional franchise part as well? We should probably make it clear this time around.

"This needs revision"
Yes.

"Robots"
"Unless the creator or singer is impossible to find" added on to the end I guess.

"Delete this"
The original idea of it meant for the base title, artist, and source fields and not for additional info as they can be adjusted for standardization or converted in either fashion.

"Regarding half-width & full-width usages of characters in the Unicode & source fields:
(Brought up to me by S o h)
Special characters should retain their original full-width/half-width characters in the Unicode fields. An exception to this is when it used for additional complimentary info like the CV section or mix descriptors. Improper usages can result in errors while searching. https://osu.ppy.sh/ss/10623085
Example using "カラフル。(Extended edit)"
The period cannot be substituted for its counterpart. "カラフル.(Extended edit)" is not acceptable.
The parenthesis may be either half or full-width. "カラフル。(Extended edit)" is acceptable.

Original width usages should still be prioritized in the unicode field when possible. "

My original post. Let's aim for this instead. Standardization edits like vs. and whitespaces should also be included in this as exempt.

I believe that covers what you have there. I need to read the updated draft when I have more time, so if something was already applied or a better method was used, then awesome.
Nevo
But tv size markers are official and aren't a preference set by specific websites and what not. Making it so we have to change every song with a marker picked from who ever released it just so it's standard aming maps is like forcing all maps to be capitalized in the same way, since then it's standard among all maps.
Lasse
just another small addition:
since CV: and vs. are already being standardized, I guess feat./ft./featuring/... should also be standardized to follow the same thing (probably feat.) for all of those?

I'd personally still prefer to just use official sources for all of these things, but that doesn't seem to be an option, so at least standardization should be applied more consistently
Okoayu
note to self: the trailing whitespace thign is about the romanised fields mostly, if the feat. is followed by full width characters it doesn't really matter because languages like japanese don't really do spaces anyways.
DeletedUser_1981781
After looking at this DQ I noticed the "Various Artists" rule is very vague especially for anime songs.

Since we already have a rule for doujin circles, it would be good to include that in the cases where the artists are five or more but are obviously related for being part of the same cast those should be credited as "show cast" or any variation to it, in the example above it would be "Toji no Miko Cast - "

salu2
Vulkin
What about the casts that have an internal band name? Like for "Zettai Tokken Shuchou Shimasu!" there are just 5 singers ( Airi Totoki, Akane Hino, Aiko Takamori, Syoko Hoshi, and Yuko Hori , omitting the CV's), and thus could use Various Artists, But the group of these singers is "Zekkenzu!", Under that case, should one put just "Zekkenzu!" at Artists? All the singers with their CV's? or Various Artists?
DeletedUser_1981781

Vulkin wrote:

What about the casts that have an internal band name? Like for "Zettai Tokken Shuchou Shimasu!" there are just 5 singers ( Airi Totoki, Akane Hino, Aiko Takamori, Syoko Hoshi, and Yuko Hori , omitting the CV's), and thus could use Various Artists, But the group of these singers is "Zekkenzu!", Under that case, should one put just "Zekkenzu!" at Artists? All the singers with their CV's? or Various Artists?
That's a nice one, Idol Units/Broadcast Units are a common thing in the voice acting scene nowadays. I'd go for "Unit Name - Song Name" and adding CVs and characters' names in the tags.
Serizawa Haruki
Wouldn't it make sense to add a rule about (Short Version) for visual novel/eroge openings to be consistent with the TV Size rule?
I'm pretty sure it has been brought up before but the thread seems dead so yeah
Okoayu
OK GUYS

FIXED, HERE'S THE CURRENT STATE OF MATTERS:
https://gist.github.com/Okorin/15b12769091f4f6750625a5572245ea8

PLEASE LEAVE LIKES, COMMENTS AND SUBSCRIBE TO MY YOUTUBE CHANNEL ON THE DOCUMENT, WE WILL ADDRESS THE REMAINING COMMENTS IN TWO WEEKS

IT'S A COMPLETE REWRITE, THIS IS THE WHOLE METADATA SECTION AFTER ALL IS SAID AND DONE, IF YOURE MISSING ANYTHING, PLEASE REFER TO THE CURRENT WORDING ON THE RANKING CRITERIA
_handholding
Official Sources must be used as references for metadata unless none are available, in which case use what is most common and recognizable.
#subjective
Xinnoh
is it just me or would having an all caps artist/title song be strange to enforce having a lowercase vs

eg
USAO VS. CAMELLIA - ULTRA HYPE WUBS

USAO vs. CAMELLIA - ULTRA HYPE WUBS

the lowercase really changes the way it's read, something like that would kinda go against the artist's intentions
ya i posted not on the document but it's kind of a separate point from what's there
Topic Starter
Noffy
Mostly everyone ended up literally posting on the gist linked by okorin okoratu , https://gist.github.com/Okorin/15b12769091f4f6750625a5572245ea8 discussion can be seen in the comments section.
Posting there kinda made this thread look super dead when it wasn't so let's not do that again

A Pull Request for the changes is now currently open on github
https://github.com/ppy/osu-wiki/pull/1847
Okoayu
https://github.com/ppy/osu-wiki/pull/1847 :clap: emoji goes here

we done with this for now boys
further amendments go into another thread
Ayana Hana
Please sign in to reply.

New reply