Thanks for all the replies to my post!
First and foremost, I'm sorry if I appeared to be making demands as if this was a service I had paid good money for. I recognize and deeply respect all the effort that has gone and is going into the librivox project for no profit at all. I merely wanted to point out a few areas where improvement (imho) is possible/desirable. Unfortunately, right now I don't have the time to look into the specifics of several of the issues that arose, but I hope to remedy that during the next few days. I'll answer what can be answered relatively quickly (though I have just spend over half an hour in trying to find MAC software
)
hugh wrote:
1. No file sizes given with the links.
our process is this: individuals submit their files to book coordinators, who then submit to our (volunteer) catalogger [...], who then must check files and upload to archive.org, then import, add relevant metadata etc. This is a thankless, tedious, and time-consuming job as it is. [...] We are currently building a more automated system [...]
This is a process I'll obviously have to look into more closely if I want to say anything serious about it. However, what are you expecting from the automated system? It shouldn't be to hard to whip up a simple & basic Python script that takes an uncompressed/losslessily compressed input file (wav, flac, ...) and do anything that needs to be done with it. I.e., convert it to various file formats, upload it (depends on the exact upload procedure on archive.org, though), extract meta-info and even generate a first draft of the publication website. And yes, I could imagine trying to come up with something in that area, once I understand what is needed. I can't make any promises beforehand, and it definitely will not be beautiful
.
hugh wrote:
2 "Ogg" isn't really saying much.
For those in the know you are right, for most people tho, ogg IS ogg vorbis, but your point is taken. we should specify at vorbis.
I know I was being nit-picky, but in this case doing it right wouldn't hurt anyone
. You might even consider underlaying the "ogg vorbis" with a link to the xiph homepage, so people can read up on that format.
hugh wrote:speex would be fine (tho see below), tho I know none of my players (on my mac ibook) wish to play speex for whatever reason.
Mac... oh, oops. Sometimes I wonder where the Mac got it's "multimedia computer" reputation from
. In a more serious vain, there really seems to be a lack of audio players capable of playing speex for the mac. I don't have the time right now to look any further, but these are the possible solutions I've found:
The Xiph Quicktime project seems to integrate with iTunes, and is thus probably the preferable way to go. Then there's the
Java speex player, which, being Java, should run on the Mac as well. And then there's the
speex Darwinport, which seems to be part of a larger project, no idea. Maybe someone could try these and tell us of the results? Especially the ease of installation and usage would be of interest here, I think.
By the way, for windows there are loads of possible ways to play speex files, including:
- Foobar2000, a player that may be ugly in it's basic configuration, but still very powerful, versatile, and an excellent format converter, too.
- The Winamp plugin on the recommendable rarewares site. This has to be unpacked into the Winamp\plugins folder, no further configuration necessary.
- The directshow filters, which allow for any directshow based player to play ALL the ogg formats (i.e. vorbis, speex, flac and theora). AFAIK this should allow playing speex in the Windows Media Player, though because of a profound dislike for that player I've never tried it.
hugh wrote:
3. No bitrate or quality setting given for the ogg vorbis file.
[...]I don't know what bitrate archive generates for vorbis, tho we could, of course check.
It's 128kbit, too, probably for all files. This suggests archive.org uses a constant bitrate setting for ogg vorbis, which is not at all desirable. But that's not under discussion here, for obvious reasons.
hugh wrote:
Again, if a volunteer can find a way to add formats, compression, or in any way help the project, I am supportive, as long as it does not add to the already arduous, thankless, and difficult job that our catalogging volunteers take on.
As I said before, I'll have to look into the whole procedure and see how much effort it would be to come up with a very basic script that would automate at least some of the steps necessary. I can't make any promises, since if there's more to it than I'm seeing atm, I shouldn't really let my university workload suffer even more
. But so far it SEEMS such a script would only have to take an input file, compress it to various formats, add tags that can either be derived from a descriptive filename, or entered by the user, upload it to archive.org (this totally depends on the conditions of archive.org uploads, i have absolutely no idea how they work) and generate an html page containing tag derived information and download links.
And now some meat -
here are the Raven transcodes I spoke of in my first post. Both are transcoded from the ogg vorbis file@128kbit, raven_q4 is transcoded at quality=4 setting, raven_q1 at quality=1 setting. The vorbis files was 8.7MB, the speex q4 file is extremely well listenable (imho) and and 1MB, the speex q1 file is still okay to listen to and it's only a meager 660KB(!). Try them for yourselves, compare them to the downloads on the librivox page, and tell us what you think. For those interested in the gory details, I used Foobar2000 to convert the files, and the settings for speexenc were: -u --vbr --quality 4 %d.
One last issue: Did anyone ever try to create 32000 Hz mp3s instead of the usual 44100 Hz and upload those? Without testing it, I assume 32000 Hz is more than enough for mp3 voice files, and shaving off about one fourth of the frequency information should result in higher quality encodes especially at 64kbs. And yes, the speex files I transcoded are at 32KHz, resampled using SSRC before transcoding them. In Foobar2000 that's one step
.