New feature: music source

[pepak]

I wonder if it would be possible to add "music source" to the listed albums. I am thinking stuff such as:
- lossless - studio recording (e.g. http://vgmdb.net/album/2284)
- lossless - live recording (e.g. http://vgmdb.net/album/10019)
- lossy (16 kHz cutoff) (e.g. http://vgmdb.net/album/16041)
- even more lossy (11 kHz cutoff)
- MIDI render (e.g. http://vgmdb.net/album/2235)
- original sound (e.g. http://vgmdb.net/album/7492)

I guess most albums would have "unknown" value, but where the source is known, it can be very useful (e.g. for the purpose of not hunting for that $200+ CD which is bound to disappoint me due to its low sound quality).

[Efendija]

It's not so simple, lossless encodings have its variants: 44.1 kHz freq with 16 bit precision is the most common one, but there are even more... how to say... lossless versions (96 or 192 kHz with 24 bit precision, even 44.1/48 kHz with 24 bit).
Lossy encodings are more diverse: complete cutoff for mp3 at variable bitrate v0 setting starts at 19.5 kHz, for constant bitrate at 320 kbps it's 20.5 kHz... cbr at 128 kbps at 16 kHz... Compression artifacts (but not complete cutoff) is most of the time present from 16 kHz and above... AAC codec with LC (low complexity) if set to highest quality can produce no visible cutoff with very little artifacts... OGG also has very, very little artifacts if set to q10 parameter (500 kbps) and so on...
The database isn't ready yet for that kind of details (I think).

[dancey]

Devil's advocate here, but I don't think very many of us have the technical know-how to figure all that out. It's really not that important, imo. I think source (cd, vinyl, tape, digital) is sufficient enough for the database.

[Secret Squirrel]

I'd love for us to be able to catalog some more technical details like this, but we really don't have any way to find this information out for the majority of albums, other than educated interpretation of the output of some tool, and I think that in many cases that could still be guesswork. I do think we need to prepare for this kind of metadata for when we are able to start reporting details about how music is store in-game.

[pepak]

Originally Posted by Efendija

It's not so simple, lossless encodings have its variants: 44.1 kHz freq with 16 bit precision is the most common one, but there are even more... how to say... lossless versions (96 or 192 kHz with 24 bit precision, even 44.1/48 kHz with 24 bit).
Lossy encodings are more diverse: complete cutoff for mp3 at variable bitrate v0 setting starts at 19.5 kHz, for constant bitrate at 320 kbps it's 20.5 kHz... cbr at 128 kbps at 16 kHz... Compression artifacts (but not complete cutoff) is most of the time present from 16 kHz and above... AAC codec with LC (low complexity) if set to highest quality can produce no visible cutoff with very little artifacts... OGG also has very, very little artifacts if set to q10 parameter (500 kbps) and so on...
The database isn't ready yet for that kind of details (I think).

It could be something along these lines, not excatly the options I proposed. How about calling it "Sound Quality" with options such as:
- SACD
- DVD-Audio
- CD
- MP3/Lossy
- MIDI
- Original Sound

[ilef]

This kind of information can be very useful, but dancey, Efendija and Secret Squirrel make valid points:

1) If VGMdb starts to classify such data, it will have to be as accurate as possible - bitrates, encoder used, frequency cutoff[s]... if there is any...

2) The majority of users isn't educated enough about the matter, and relying on third-parties tools which might even return false positives/negatives isn't really appropriate.

Not ready yet to have such information collected and displayed.

Want to know the actual quality of an album? For the time being we have to take our risks and purchase.....It may suck, but it's also part of the fun.

[Hellacia]

Originally Posted by Efendija

complete cutoff for mp3 at variable bitrate v0 setting starts at 19.5 kHz

Slightly off the original topic now (forgive me), but I just wanted to clarify something: in the most recent stable build of LAME, 3.99 (excluding betas and whatnot), when encoding with VBR V -0, the lowpass filter is disabled, right? I encode with LAME 3.99 and I use to do 320kbps, and got a message about the lowpass filter working at about 20.5 KHz, like you said. However, now I use VBR -V 0, and I get a message saying the lowpass filter was disabled.

Back on topic now, as much as I like technical stuff, I don't think the database is ready for this kind of information, nor do I think it even has much of a place here. And, much of this information isn't discernible anyway, for some of the reasons Efendija listed. Going with your second suggestion, we're basically doing that right now with the media type field (CD, DVD, Digital, etc).

[Efendija]

Originally Posted by Hellacia

Slightly off the original topic now (forgive me), but I just wanted to clarify something: in the most recent stable build of LAME, 3.99 (excluding betas and whatnot), when encoding with VBR V -0, the lowpass filter is disabled, right? I encode with LAME 3.99 and I use to do 320kbps, and got a message about the lowpass filter working at about 20.5 KHz, like you said. However, now I use VBR -V 0, and I get a message saying the lowpass filter was disabled.

Yeah, VBR V0 is my favorite for mp3 encoding and I did try 3.99 stable and I think 3.98 is still better for V0... Here are some screens with my comments along:

(Source was lossless of course, view is zoomed on the same spot)





Upper image is CBR 320 kbps with Lame 3.98, and the one below is the same only with 3.99.... Lowpass is still 20.5 kHz for both, but it seems some more information from higher frequency range is kept in 3.99.

Mediainfo details:

Audio
Format : MPEG Audio
Format version : Version 1
Format profile : Layer 3
Mode : Joint stereo
Mode extension : MS Stereo
Duration : 3mn 9s
Bit rate mode : Constant
Bit rate : 320 Kbps
Channel(s) : 2 channels
Sampling rate : 44.1 KHz
Compression mode : Lossy
Stream size : 7.21 MiB (100%)
Writing library : LAME3.98r
Encoding settings : -m j -V 4 -q 3 -lowpass 20.5

Audio
Format : MPEG Audio
Format version : Version 1
Format profile : Layer 3
Mode : Joint stereo
Mode extension : MS Stereo
Duration : 3mn 9s
Bit rate mode : Constant
Bit rate : 320 Kbps
Channel(s) : 2 channels
Sampling rate : 44.1 KHz
Compression mode : Lossy
Stream size : 7.21 MiB (100%)
Writing library : LAME3.99r
Encoding settings : -m j -V 4 -q 3 -lowpass 20.5





These two images are V0 (upper lame 3.98, lower 3.99). In this new version lowpass filter is disabled or set at 22.1 kHz, which is the same thing, at least for source material with 44.1 kHz freq. This is bad because part of very high frequency sound data is preserved, while the data with a bit lower frequency is lost more in 3.99 than 3.98. I think like this because of the nature of normal human hearing range which is (biological maximum) 20 kHz (per channel). Higher frequencies human ear just can't register at all. Take note that 20 kHz range is if you have really, really perfect hearing, but most people around 30 years old or younger have even lower range (about 18-19 kHz). Back to this, it means new 3.99 V0 keeps some more data above the audible range, while it sacrifices more data on lower (audible) frequencies. On the other hand V0 with 3.98 has lowpass at 19.5 kHz, cutting everything above (which anyway you just can't hear), but keeps more data from 16 kHz to 19.5 kHz range which is more important.

Or to put V0 encode spectral views side by side - left 3.98, right 3.99; it's noticeable 3.99 loses more data in 16-19.5 kHz range... (I've marked it with those green lines lol)



That's why I keep doing mp3 v0 encodes with 3.98, for now. Thanks Hellacia if you had the patience to read this lmao

Mediainfo for V0 encodes:

Audio
Format : MPEG Audio
Format version : Version 1
Format profile : Layer 3
Mode : Joint stereo
Mode extension : MS Stereo
Duration : 3mn 9s
Bit rate mode : Variable
Bit rate : 294 Kbps
Minimum bit rate : 32.0 Kbps
Channel(s) : 2 channels
Sampling rate : 44.1 KHz
Compression mode : Lossy
Stream size : 6.64 MiB (100%)
Writing library : LAME3.98r
Encoding settings : -m j -V 0 -q 0 -lowpass 19.5 --vbr-new -b 32

Audio
Format : MPEG Audio
Format version : Version 1
Format profile : Layer 3
Mode : Joint stereo
Mode extension : MS Stereo
Duration : 3mn 9s
Bit rate mode : Variable
Bit rate : 284 Kbps
Minimum bit rate : 32.0 Kbps
Channel(s) : 2 channels
Sampling rate : 44.1 KHz
Compression mode : Lossy
Stream size : 6.40 MiB (100%)
Writing library : LAME3.99r
Encoding settings : -m j -V 0 -q 0 -lowpass 22.1 --vbr-new -b 32

[dancey]

I have no problem with encoder preferences or discussions therein, but everything you brought up in that post has nothing to do with the vgmdb database and everything to do with a user's personal encoding choice. That should really be moved to another thread if you want to continue it.

There's some value in adding a subcategory for the format specific information, like "SACD" as a subcategory of the CD or an RPM subcategory for vinyl, or encoder / bitrate as a subcategory for digital, but storing things like encoder settings, mode, sample rate, passthru filters used are completely worthless to pretty much everyone except the most picky of audiophiles. I cannot realistically see anyone ever coming here, looking up a digital soundtrack to a game, seeing that a soundtrack is only available in mp3 vbr and saying "Well, I'm only going to buy it if it's lossless!" and then not buying it.

It would require an extraordinary effort to gather that information about current and future releases. Someone would need to buy it (not that uncommon), then find software to analyze it, then provide the information. Some of the information may not even be available (encoder settings might not be).

What it really comes down to is it's entirely up to the content provider on how to encode it and storing that information in this database isn't going to add any value or help people make informed decisions beyond what they already care about (format, bitrate, compression).

[Efendija]

Originally Posted by dancey

I have no problem with encoder preferences or discussions therein, but everything you brought up in that post has nothing to do with the vgmdb database and everything to do with a user's personal encoding choice. That should really be moved to another thread if you want to continue it.

I was just answering Hellacia's question and I'm finished... I'm aware this was off-topic, btw.

[Efendija]

BUT, it also shows who the fuck would go through such analysis just to make a vgmdb entry, so in the end it's not completely off topic, dancey

[TerraEpon]

Of course, there's only one video game SACD in existence AFAIK -- the DQVII Symphonic Suite.

[Cedille]

I didn't know we could search by SA-CD, but Aquaplus seems to like it anyway.

http://vgmdb.net/search?do=results&id=130753

[pepak]

Originally Posted by Hellacia

Going with your second suggestion, we're basically doing that right now with the media type field (CD, DVD, Digital, etc).

No, we are not. CD can still be made from lossy source (which is in fact the case with all Ubisoft titles that I know of), or they can be made from MIDI source (which is the case with some rather expensive CDs, such as Startrail or [I am told] the Daggerfall soundtrack, The Music From Tamriel). Especially in these cases it makes sense to warn people beforehand of what they are getting for their $100+.

[Efendija]

^That's correct, CD/ Digital etc. doesn't necessarily describe the quality. There are bootlegs (and even official releases) which are made completely from lossy sources; digital distribution doesn't always mean the files are lossy (like Mass Effect 3 N7 Collector's Edition OST, it's digital download, but it's lossless); there are cases when CD release has some mix of lossless and lossy tracks (Borderlands OST) and so on. Current vgmdb category only states the distribution method, not the quality of tracks... But it should stay like that - who wants to pay 100+ bucks for a pontentially lossy mastered tracks on a CD isn't probably going to ever fire up any spectrogram analysis software and such, so no point really...

[suicider]

Wouldn't it be an option to allow the addition of the recording-source to the notes field for albums where the source is known?

[Datschge]

The only info I could possibly want is if the original source is already clipped to the hell and back (i.e. loudness war). Those I'd mark with a big fat warning to potential buyers. Everything else is a matter of taste and often too technical to really constantly qualify for the whole database.

[Phonograph]

instead of having some new categories like Lossless, Lossy, etc.
you could simply warn people in the notes or forums that the album isn't made from a good quality source