Audiosurf and the Loudness War

[UncleChuckle]

If you don't know what the Loudness War is, read the following link.

http://en.wikipedia.org/wiki/Loudness_war

Dynamic Range Compression.

The reason it just popped into my head is a song I just did I got a lowish score on. Compared it to the higher scores, and saw a 4 second time difference no biggie. But looking at the "map" of the level, while the same peaks and troughs are there, they are much more emphatic on the other higher scoring ones. (My score is low because I hit grey blocks. Not trying to blame loudness for my crap performance. Try playing Audiosurf while suffering from the flu and see how you do.)

The following page has a waveform comparison comparing original release material (like U2's "With or Without") with "remastered" where the dynamic range compression has been cranked up.

http://www.rollingstone.com/news/story/17777619/the_death_of_high_fidelity/print

So just wondered what sort of potential differences in scores it would lead too if someone played the original "Joshua Tree" version of "With or Without You" vs the remastered (and some, myself included, would say destroyed) version.

[UnknownAssassinX]

I'm curious about this, now, too. I'd never heard of it up until now. (I'm sick with what seems to be the flu. Everyone in my city's got the damn thing too, it's so bad that there was fucking three cans of soup left at the grocery store when I went to pick some up.)

I would suggest you do a test. Get the remastered, get the original. Change the ID3 tags of both, then play both and compare. I could help with this if you'd like, though I have no songs that have been remastered, as far as I know.

[Psy-T]

I'd wager the remastered versions would offer more points, not to mention a noticeably different track shape.

[Faxmachinen]

That brings up an interesting point - how does Audiosurf identify the songs?
The most obvious and safest approach would be an MD5 of the file, but that would mean that different formats and compressions, even those that reproduce the audio data virtually identically (e.g. CDA and FLAC) would have different scoreboards.
I would really hope you're wrong about it just using the ID3 tags though. It would hardly take a hacker to cheat their way to the top of the scoreboard then.

[Phantasmagoria]

I've noticed this kind of track difference on a number of songs. Whether due to dynamic range compression or ordinary lossy software compression, or both, I don't know. Here's one of the more significant examples I've come across (Away with the Faeries, by Inkubus Sukkubus):

[UnknownAssassinX]

Phant: That, honestly, sucks. I actually have an idea of how to fix this though:

The highscores could be programmed to detect LARGE differences between the graphs, and then tag them as seperate versions if it turns out that there is a large difference. Wouldn't be hard to do, kindof heavy on bandwidth but it looks like the server already gets sent a graph so it wouldn't be anything more than it already deals with.

[Melodia]

That brings up an interesting point - how does Audiosurf identify the songs?
The most obvious and safest approach would be an MD5 of the file, but that would mean that different formats and compressions, even those that reproduce the audio data virtually identically (e.g. CDA and FLAC) would have different scoreboards.
I would really hope you're wrong about it just using the ID3 tags though. It would hardly take a hacker to cheat their way to the top of the scoreboard then.

Sorry to disappoint you, but that's in fact exactly what it does. It looks at the title and the artist (and JUST those two) in the ID3 tag. So not only is it easy to cheat, but it ALSO means that many people will play the exactly same thing and be on different scoreboards because they have things tagged differently -- often completely correctly.

In a bit I'll be making a nice long detailed post on the issue, and hope others join in...


-Lala-

[Tim Allen]

I'm curious about this, now, too. I'd never heard of it up until now. (I'm sick with what seems to be the flu. Everyone in my city's got the damn thing too, it's so bad that there was fucking three cans of soup left at the grocery store when I went to pick some up.)

I would suggest you do a test. Get the remastered, get the original. Change the ID3 tags of both, then play both and compare. I could help with this if you'd like, though I have no songs that have been remastered, as far as I know.

hope you get well soon bud.

[UnknownAssassinX]

I'm curious about this, now, too. I'd never heard of it up until now. (I'm sick with what seems to be the flu. Everyone in my city's got the damn thing too, it's so bad that there was fucking three cans of soup left at the grocery store when I went to pick some up.)

I would suggest you do a test. Get the remastered, get the original. Change the ID3 tags of both, then play both and compare. I could help with this if you'd like, though I have no songs that have been remastered, as far as I know.

hope you get well soon bud.

Thanks, Tim Allen. Coming from you, I'm sure to get better very soon.

[Slanzinger]

It ALSO means that many people will play the exactly same thing and be on different scoreboards because they have things tagged differently -- often completely correctly.

Aye, this has happened with me on "I Don't Feel Like Dancin'" by Scissor Sisters - I think there's two of us with the apostrophe on the end, and a whole full set of three boards for ones without, even though on the game itself the names appear identical (It doesn't seem to see apostrophes)

Maybe if the game disregarded punctuaction like that, the scoring would work better? (so that I Don't Feel Like Dancin' merges with I Don't Feel Like Dancin and I Dont Feel Like Dancin' [and so on ad nauseum])

[Dark-Sonik]

I've noticed this kind of track difference on a number of songs. Whether due to dynamic range compression or ordinary lossy software compression, or both, I don't know. Here's one of the more significant examples I've come across (Away with the Faeries, by Inkubus Sukkubus):

I know another good example: Got two versions from Killswitch Engage's "My Curse" where one of them is the usual one you can find everywhere and buy, and another one which someone recorded out from Guitar Hero 3 (that version has a silent shouter and by 3 seconds longer ending), but the recording is that f**ked up that the song is more quiet than the normal one, thus making it slower and giving less more points.

And since most people have downloaded songs from many sources - such as sometimes ripped versions, badly converted versions or also a clean version -we get different scores, so, as said, it'd be good to record the MD5- and/or the checksum of the files and take them to the scoreboards, just so we get some scores seperated and not some people yelling "Cheater!" because someone else with a louder version got more points... (Except that person boosted it on purpose)

[UncleChuckle]

Wish I had something in original and remastered form to test, but off the top of my head I don't have anything.

Though I do wonder if the original (where there is more dynamic range to the sound) would create bigger peaks and troughs. That would make sense I think.

[Pwntastic]

I do like the md5/checksum idea. Could have it autoflag anybody with alternate checksums, especially those out of a certain range.

Then you just need to rearrange highscore tables to allow for various bitrates / encoding schemes and you'd be close to set