the foobar2000 plugin: "If the maximum peak difference between tracks is less than 0.3 dB then the DR value is computed with all tracks processed as a single file (Album Mode). Otherwise the DR value is computed as the rounded value of the average of all single tracks (Song Mode)."
2012-08-23 23620, 2012
hawke_1
I suppose I should put those on the wiki
2012-08-23 23608, 2012
kepstin-work
so it looks like the official way is to just average the dr numbers for the album in most cases... I guess.
2012-08-23 23643, 2012
hawke_1
Is it?
2012-08-23 23614, 2012
kepstin-work
should download the official tool and run some comparisons to make sure
2012-08-23 23630, 2012
kepstin-work
but I assume most people use the foobar2000 plugin, which does what I quoted above.
2012-08-23 23630, 2012
_5moufl
or take the median value
2012-08-23 23656, 2012
_5moufl
which probably makes more sense
2012-08-23 23613, 2012
kepstin-work
if it turns out that there's no real way to calculate an album value (like with album replaygain) we could just not submit one, and let the server deal with it.
2012-08-23 23614, 2012
hawke_1
The foobar2000 plugin is semi-official, btw
2012-08-23 23606, 2012
kepstin-work
hmm.
2012-08-23 23628, 2012
kepstin-work
well, calculating album mode would probably be a good addition
2012-08-23 23618, 2012
kepstin-work
unless there's something clever that can be done, it would mean that you can't parallelize album calculations
2012-08-23 23625, 2012
kepstin-work
(replaygain is annoying like that too)
2012-08-23 23653, 2012
hawke_1
Right
2012-08-23 23603, 2012
hawke_1
Though I think it’s fast enough that that’s not a big deal.
2012-08-23 23628, 2012
kepstin-work
can always do albums in parallel :)
2012-08-23 23651, 2012
hawke_1
Yep.
2012-08-23 23602, 2012
kepstin-work
hmm. actually, this works with a histogram.
2012-08-23 23617, 2012
kepstin-work
the histogram could be calculated for each file then merged.
2012-08-23 23644, 2012
ianmcorvidae
unless the buckets differ between files, I guess
2012-08-23 23648, 2012
kepstin-work
it's trivially parallelizable
2012-08-23 23650, 2012
ianmcorvidae
but if there are fixed buckets then it should be fine
2012-08-23 23655, 2012
kepstin-work
Each RMS and Peak value is entered into a histogram with 10,000 discrete values ranging
2012-08-23 23657, 2012
kepstin-work
from -100 dB to 0 dB in steps of 0.01 dB.
2012-08-23 23659, 2012
hawke_1
Pretty sure they’re fixed.
2012-08-23 23615, 2012
kepstin-work
you could even parallelize it within a file if you like :)
2012-08-23 23631, 2012
kepstin-work
that that it would save you much work
2012-08-23 23642, 2012
kepstin-work
(given that file decoding is probably the hardest bit)
2012-08-23 23604, 2012
hawke_1
It uses subprocesses to decode to wav, btw
2012-08-23 23628, 2012
hawke_1
So that could almost certainly be improved.
2012-08-23 23618, 2012
hawke_1
It’s really not a very well-written program.
2012-08-23 23627, 2012
kepstin-work
hmm. the histogram only needs about 40KiB ram per file, so there shouldn't be any memory issue for reasonable numbers of files
2012-08-23 23659, 2012
kepstin-work
I know, I'm tempted to write the tool myself, or start from the C version that you found.
2012-08-23 23615, 2012
CallerNo6
demosdemon: are you saying that a soundcloud track needs to start with a capital letter? (not arguing, just wondering)
2012-08-23 23636, 2012
YukiS joined the channel
2012-08-23 23606, 2012
hawke_1
I have no idea if the adiblol program is any better.
2012-08-23 23611, 2012
hawke_1
But at least it uses ffmpeg to decode
2012-08-23 23655, 2012
YukiS
I've enabled the move featured artists to track title plugin but the renaming still treats the album as various artists, any workaround?
2012-08-23 23608, 2012
demosdemon
CallerNo6, after some digging, I'll retract that because I've found some tracks with lowercase titles. I remember a time when soundcloud would enforce their own capitalization schemes but that's apparently changed
2012-08-23 23633, 2012
CallerNo6
Just the same, good to know. I'll keep it in mind.
You voted "no" and said "the HIB version is the same as the bandcamp version" which is untrue.
2012-08-23 23619, 2012
intgr
That's like "I'm voting No on your edit because the world is flat"
2012-08-23 23627, 2012
hawke_1
Actually no
2012-08-23 23633, 2012
intgr
Thank you :)
2012-08-23 23638, 2012
hawke_1
The bandcamp one has 23 tracks too.
2012-08-23 23649, 2012
intgr
Huh?
2012-08-23 23600, 2012
demosdemon
no, I only see 21
2012-08-23 23600, 2012
kepstin-work
a lot of bandcamp releases have hidden tracks
2012-08-23 23624, 2012
kepstin-work
(they aren't shown on the website, but are included in the download)
2012-08-23 23636, 2012
CallerNo6
when did life become so complicated?
2012-08-23 23637, 2012
hawke_1
2bonus tracks included!
2012-08-23 23642, 2012
intgr
:/
2012-08-23 23657, 2012
kepstin-work
you might be able to get info about them from the api, I dunno.
2012-08-23 23611, 2012
demosdemon
oh, yeah
2012-08-23 23613, 2012
demosdemon
there it is
2012-08-23 23627, 2012
intgr
Holy crap
2012-08-23 23612, 2012
demosdemon
CallerNo6: when the first homosapien walked the earth :P
2012-08-23 23611, 2012
ijabz joined the channel
2012-08-23 23615, 2012
demosdemon
I think this whole process is a perfect example as to why fast tracks shouldn't happen btw. The faster you want something the happen, the more likely there will be mistakes.
kepstin-work wonders if nes will support reverting merges
2012-08-23 23656, 2012
demosdemon
it's supposed to
2012-08-23 23624, 2012
kepstin-work
iirc one of its goals was easily reverting data-loss edits :)
2012-08-23 23609, 2012
v6lur
ianmcorvidae: "Username field is required", "Please enter a password in this field", "Please enter a password confirmation" are untranslatable on /register
2012-08-23 23620, 2012
v6lur
(sorry, i'm too lazy to properly file bugs)
2012-08-23 23647, 2012
CallerNo6
ha! from the older edit demosdemon linked to, "Follow up: I asked DVA on the issue over three weeks ago, no response so far... "
2012-08-23 23617, 2012
CallerNo6
So much for my idea that musicians love to answer emails.
2012-08-23 23622, 2012
demosdemon
heh
2012-08-23 23653, 2012
ijabz_ joined the channel
2012-08-23 23649, 2012
_5moufl
CallerNo6: I missed that, I sent him a pm on fb like 1h ago
2012-08-23 23633, 2012
hawke_1
kepstin-work: The C version of dr_meter is very very fast. As you might expect.
2012-08-23 23608, 2012
ianmcorvidae
most of the heavy lifting the python one does should be C anyway, isn't it mostly using numpy?
2012-08-23 23637, 2012
hawke_1
ianmcorvidae: I expect most of the python one’s time is spent writing the file out to wav… :-/
2012-08-23 23647, 2012
ianmcorvidae
heh
2012-08-23 23649, 2012
ianmcorvidae
yeah that might do it
2012-08-23 23658, 2012
hawke_1
…yeah… :-/
2012-08-23 23602, 2012
CallerNo6
_5moufl: they maybe they'll answer, so that we'll leave them alone.
2012-08-23 23603, 2012
ianmcorvidae
which is to say, though, the detriment of the python one isn't python
