A Review by Dan Wallach <dwallach@cs.rice.edu>
January 21, 2006
I have over 1200 CDs. I sent MusicShifter half of my collection and found that they weren't anywhere near the painless experience that I'd fantasized about, but then my music collection can get quite obscure and I'm pedantic about correct metadata. I was happy enough, though, that I decided to send MusicShifter the other half of my collection.
Overall, MusicShifter only really blew it on three out of 630 CDs. The ripping quality was perfect, but the tags were entirely empty. An evening of dorking with FreeDB and iTunes was enough to solve the brain damage, for the most part, although I've seen enough tag weirdness that I'm going to need to read everything over carefully to see if there are any other surprises lurking.
I have an 8-month old daughter who may well decide that my CD collection would make great chew toys once she can reach them. I previously ripped maybe half of my collection using a variety of different tools and varying MP3 quality settings over the years. The unstable rack of CDs needed to go away, pronto. I didn't, however, have the time to spoon-feed everything into my machine. After much web surfing and soul searching, I decided to re-rip everything with a commercial service in Apple Lossless format (digression on format choice, below). I ultimately decided to use MusicShifter both for their (relatively) reasonable price, and for their claim of superior metadata quality, including embedded album art.
I shipped the first batch of 600-or-so CDs to them in August 2005. The whole packaging experience is first rate. You pay. They send you a giant box with all the packing materials. You load your CDs into 100-CD plastic spools of the sort that normally contain blank CD-R's purchased in bulk. Those spools are then wrapped in sheets of foam, placed into individual boxes, and then collected into a larger box for shipping. It's hard to imagine any better way of doing it.
I opted for the cheaper DVD-R option, rather than paying the extra fee for sending them a USB hard drive. Stupid me. Ripping 50+ DVD-R's isn't my idea of a good time, either, plus I discovered that not all DVD drives are created equal. My PowerMac G4's "SuperDrive" (an early Pioneer DVD-R burner) had all kinds of errors, even though MusicShifter uses DVD-R media. My wife's recent-vintage Dell laptop, with an internal DVD-reader/CD-burner, has a habit of overheating and giving up. I ended up using my work machine, with a Plextor burner, and copying everything to a USB hard drive to carry home. Next time, I'm just sending along a USB hard drive. (In the end, everything goes onto my home RAID box, so it should be reasonably safe against minor catastrophies.) Also, one of the 50 DVDs had a hard error that corrupted exactly one file. I dug the original CD back out and re-ripped it.
Moral of the story: spend the extra bucks and send along a USB hard drive.
MusicShifter claims to have higher quality metadata than GraceNote, and that's largely true, particularly with the album art, which you can't get from the free services (although I've heard good things about the Windows-only shareware iArt). Still, when things went wrong, they went very strangely wrong. Example: I had them rip several Timbuk 3 CDs. All but their most popular album (Greetings from Timbuk 3, with their hit song "The Future's So Bright...") came back with empty song names and no artist listed. However, they had the correct album name and the correct artwork! GraceNote or FreeDB have the proper data, so it's not like MusicShifter couldn't have gotten it right. The same thing happened with Donald Fagan's The Nightfly -- hardly an obscure album. On the flip side, they got some really obscure albums right, including three out of four 3-inch CD samplers I got back in 1987/1988. They messed up my Metallica One single in the same way as Timbuk 3, but they gave my Simple Minds single the correct song titles, albeit with the wrong year. Ahh, well. Also in the Timbuk 3 category was a Rhino "New Artist Sampler" from 1988, a 1999 Leftism "Bonus CD", and a 1980's "Guitar Speak 3" compilation.
Luckily, I had already ripped many of these, so I was able to bulk copy the metadata from my original tracks to the new ones using Doug Adams' fantastic Copy Tag Info Tracks to Tracks script (which I hacked to also copy the BPM and comments fields; don't get me started on how ugly AppleScript is to hack). Verifying whether I prefer my own tags on my existing MP3s or whether I prefer the MusicShifter tags on the new Apple Lossless files is something of a manually intensive process, so I'm doing it a bit at a time to make sure that everything is just right.
Three CDs came out as completely unknown -- no tag data at all. Did I recognize the tracks? Vaguely. Back in the day, I bought a lot of cheap stuff from the used CD bins at Amoeba and now I'm paying the price. After hours of effort, I've identified them as:
If your CD collection has odd releases from the dawn of the compact disc, you may find similar issues. (For what it's worth, all three CDs appear in FreeDB, so MusicShifter should have been able to get these tags from there, if nowhere else.) In email, MusicShifter claimed that they have now revamped their process. If a CD comes up for which they can find no metadata, they will segregate it from the stack, allowing you to at least know which CDs went unrecognized without needing to do a painstaking search.
MusicShifter claimed, in an email to me, that they regularizing the capitalization of song titles, versus the wild west of crap on FreeDB and GraceNote. Again, that's proved largely true, although I've found obscure counterexamples (e.g., Musician Magazine's A Little on the CD Side, Volume 3 -- probably late 80's), where only the first letter of each song title was capitalized. Something similar happened with We're From Texas (And You're Not), a late-80's obscure Texas punk collection. I'm guessing that when their "superior" data source has nothing, they sometimes fall back to FreeDB or GraceNote, which is definitely better than nothing. Given that these cases are somewhat unusual, it would be nice if MusicShifter could put the extra effort into polishing the tags.
Consider multi-year greatest hits collections. With MusicShifter, you tend to get the year that the collection was published (useless) rather than the year when each track was originally released. Where I've previously ripped the CD and fixed this sort of thing, I can use the aforementioned AppleScript to migrate the years over. Otherwise, too bad.
Remember that annoying trend in the early 1990's where bands thought it was cool to have a "secret" song at the end of the album, often after a bunch of empty, silent tracks? MusicShifter happily rips each of them. It would be nice if there was an option to say "really, just skip that crap."
For the Afro-Cuban All Stars' Distinto, Diferente, all but one track came back with the artist labeled "Juan De Marcos' Afro Cuban All Stars", with the other labeled the proper "Afro Cuban All Stars". Likewise, for A Toda Cuba Le Gusta, everything was properly labeled "Afro Cuban All Stars". I was kinda hoping for more consistency.
For Month Python's The Final Rip Off, the number of actual tracks is more than those advertised on the back of the packaging. They inserted a five-second blurb, at one point, and did other assorted random things you'd tend to expect from the Pythons. As a result, several of the tracks on the end were labeled with the unhelpful title "Track 24" and so forth, and no artist. Bother. Similar broken behavior occurred on another ten or more CDs, such as Cream's Fresh Cream; my CD has thirteen tracks, while the "normal" CD apparently only has eleven. With some FreeDB queries I was able to sort all of this out, but MusicShifter should really have done this as part of their standard service.
I just got back the USB drive with the results of ripping the other half of my CDs. They included a letter saying that questionable CDs are placed on the top of each spool with various colored dots that mean "we couldn't find the metadata" or "this disc was too scratched up to rip". That's good, although the same issue with tags appears to have happened again. Out of 600 or so CDs, two were completely unknown (although thankfully both CDs were tagged with blue dots and were easy to find) and another twelve had the proper album title and such, but no song titles. Many other CDs had missing artist or titles for some but not all tracks. This is somewhat excusable for obscure items (particularly some cheap Polish-label jazz CDs I picked up in Germany), but many of these issues could have been flagged and fixed as a matter of quality control. Consider Jazz Samba by Stan Getz and Charlie Byrd, a recently released Verve "Master Edition", so it's hardly an "obscure" album. The first four tracks were labelled only with a track title ("~dmcin") and nothing else. The second four tracks were fine. The files, on disk, however had the proper titles as the file names of every file. MusicShifter simply screwed these tags up for no apparent reason.
After an afternoon of effort, I largely cleaned up the mess. Still, it will take much longer to convince myself that I got everything right.
When you can rip any way you can imagine, what should you choose and why? MP3 is the obvious first choice. It saves on space and it's playable anywhere and everywhere. However, it has one very particular problem. If the lossy compression was too lossy, there's no knob you can twist or program you can run to get that quality back. These days, "audiophile" MP3 ripping seems to be done at high variable bitrates (around 200Kbps), which to my ears is indistinguishable from the original CD data in every case I've ever tried.
Well, what about newer formats, like Apple's AAC or Microsoft's WMA? It's all about support. AAC won't work in a large number of devices. Similar issues abound for WMA. Even if they are truly smaller than MP3 for an equivalent quality, they're just not useful formats if you want to put your CDs away and never need to see them again.
Lossless formats (including Apple Lossless, WMA Lossless, and FLAC), however, are a completely different animal. You're getting the original data from the audio CD. The compression, such as it is, is entirely reversible. This means you can transcode from the lossless format to your favorite lossy format, whatever works best for your portable music player or whatnot. If I want to fit a larger chunk of my collection into my laptop for DJing, I can make the quality/space tradeoff myself, and redo it if I'm not happy. Sure, my computer may need to run for a couple days straight to convert every single file. So what? Furthermore, I've got a big RAID file server at home, so I don't have to worry so much about the space issues for storing the lossless files.
Okay, which lossless format should I choose? The answer depends on what tools you like to use. I'm a big fan of iTunes. I particularly like it for its ease of doing massive editing tasks on metadata (e.g., find all songs with a title that looks like Take the 'A' Train and change them all to have one common spelling and capitalization of the song title). If it's iTunes you love, then it's Apple Lossless you should choose. On the other hand, Apple Lossless doesn't work with my SlimDevices Squeezebox or RioCar, both of which support FLAC out of the box. (UPDATE: the latest SlimServer release transcodes from Apple Lossless to uncompressed audio, on the server side, allowing a Squeezebox to play it.) However, neither of those needs lossless audio. The Squeezebox is doing ambient music for the kitchen, not high-fidelity music in the home theater. Likewise, the RioCar is in my car, where I want a bigger library and don't need high audio quality. For my home theater, I'm toying with getting an Apple AirPort Express, which supports Apple Lossless, but I'm more likely to eventually get a home theater PC of some sort, which can run iTunes and thus support Apple Lossless. Put it all together, and Apple Lossless is the right format for my needs. Obviously, if you use different tools, or have different gear, then your preferred format may be different. The joy of any lossless format is that you can, at least in principle, convert from one to another with absolutely no loss of quality.