Slashdot Mirror
Ask Slashdot: What's the Best MP3 Encoder?
syd asks: "I'm wanting to convert my CD collection to MP3, and I'm looking for the best MP3
encoder to do the job. The most important factor is the sound quality
of the encoded files. Other concerns are cost, platform, and speed of
the encoder. However, I'm only going to encode them once, and I'm
going to listen to them fairly often, so I'd rather have a slow
encoder that sounds good. I would prefer to use Linux, although I
would be willing to reboot into Windows if necessary. If anyone has
any pointers to some real numbers, that would be most helpful."
A fair enough question. What do you all think?
---
I agree with most of what you say except for the "gimmicky variable bit rate option." Variable Bit Rate (VBR) is a Good Thing(tm). It allows the encoder to use fewer bits for unimportant sections (like complete silence) and more bits for complex portions using many frequencies. This makes for a better sound quality given the same filesize.
The only disadvantage I see with it is that it messes up some crappy mp3 players, but that's a problem with the mp3 player, not the encoder.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
Although most of the focus centers on the mp3 encoder, people often forget that the CD ripper is important as well. While bad encoding may be annoying, skips in mp3s are infinitely more annoying.
For rippers that work on IDE CD-ROM drives, I've personally found all the ones integrated with encoders to be crap. On my (fairly bad) CD-ROM drive, they almost always produce mp3s with skips in them. Worse yet, they don't do much of any error checking to let me know of this.
So I use CD-R-Win (for Windows). It's designed for CD-to-CD copying with a CD-R drive, but can also do CD-to-WAV ripping with any CD-ROM drive. It actually does extensive error checking, and NEVER produces a file with a skip, no matter how small, in it. You have an option of having it abort in an error, or having it retry reading the skipped area a certain number of times to correct the error. To the best of my knowledge, no other ripper has this option. With every other ripper I've tried, I've gotten at least one or two mp3s with skips in them, while cdrwin is 100% perfect.
That said, I'd prefer a ripper integrated with an encoder and CDDB lookup. However, the ripper must be 100% perfect, and NEVER produce a skip. Does anybody know of a ripper other than cdrwin that can do this? So far I've been unable to do so, though I'll continue looking.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
I've tried CDex, but I haven't been able to get it to rip without skipping at least once per song. It is better than most of the other rippers though, since it at least tells you when it had skip problems. Unfortunately, it doesn't actually fix those skip problems, which is what cdrwin does (and what I'm looking for).
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
Wow, didn't realize it had progressed that far. I'm using CDex v0.15 beta 4, which could be why it's not performing as well as you described =)
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
Grip + LAME is my choice. The auto rip and eject features are nice for bulk ripping.
I share my rip directory over NFS, and use a perl front end to lame to dispatch encode jobs over the network so it can encode several tracks in parallel. With enough boxes, it doesn't matter how slow encoding is.
I've had disappointing results with Fraunhoffer. The authors admit that they wrote it with low bitrate voice recordings in mind.
In general i suppose it isn't bad, but it's joint stereo I dislike. It's an imperfect way of doing things.
Some of the bands i listen to sometimes use weird qsound-alike spatial effects when they mix the album. fraunhoffer slaughters those and makes this swooping noise. it's irritating. Tracks of note are LPD's "10th Shade" and Japan's "Gentlemen Take Polaroids"
I wouldn't say bladeenc is perfect. But, well, nothing is. I use bladeenc and it isn't half bad.
From all reports, Xing's encoder uses a really cheezy approximation of the codec. That would be in keeping with Xing's history in the market - their mpeg video encoder creates mpeg videos that are comprised entirely of index frames.
This is just like television, only you can see much further.
On Sarah McLachlan's Surfacing album, try encoding the very last track (I forgot what it's called) with different encoders. It's a really difficult song for MP3 encoders to get right, esp around 23 seconds into the song when the cello starts. Every encoder I've used so far screws up the cello at 128kbs, but you can compare which encoders handle it better than others.
Make sure you listen to it with headphones on instead of speakers. If the encoder is sub-par in quality, the cello will have some kinda wobbly sound
Do a listening test of Xing @ 128kb/s vs the real CD. If you can't tell the difference / don't care about the difference, use it, since it's the fastest.
Personally, the quality bugged me, and I used a program with the lastest Fraunhofer encoder in it (which does not have the frequency cap on it). There are a number available, but unfortunately they're all windows-based. It was *very* difficult for me to tell the difference between the original source file and the files encoded at 128 with the high quality setting turned on.
Of course, using this setting, Fraunhofer encodes at less than 1x, so it's pretty essential to set up some sort of batched mechanism where you can rip and fill your HD with unencoded wavs and let the encoder chew on them while you're sleeping. Otherwise, you'll be spending a LOT of time in front of the box switching out CD's. Actually, you will anyway, but you can get a lot more rips done in a short amount of time if you're not waiting for the encode.
As far as CD-ROMs go, I can't recommend the ASUS 40x drive more highly- it rips consistently at 5x-10x without a single error that I've heard so far in my 200+ CD collection. There are faster drives out there, but they don't come as cheap- about $50. Well worth the money, especially if you consider how much time you will spend encoding a collection of any size.
This is a valid question anyone who encodes should consider. I currently have over 40 mp3 CDs worth of my collection that I've encoded. I've maxed out a 20 GB HD and will be purchasing another to put the rest of my mp3s on my server.
I encoded a large portion of my collection using Windows, specifically Audiocatalyst. It works really good most of the time. However, it did not work great with all my drives. My HP 8100i CD-R is the best drive I have to rip with. My DVD Drive and older CD-Rs would have too many skips. In going back and listening to my collection, I still run across songs that have blips in them, though this is rare. Audiocatalyst rips and encodes so fast, you don't have time to listen to all the songs prior to burning a CD, so I expected a bit of this. Of course, when you rip under windows, don't expect to be able to use it for anything else, unless you love reading hex on a blue background.
I am currently using Linux to rip and encode, and I have much better results. My CD-R is still the fastest drive to rip with, but I can rip with a 2nd gen DVD, and a 12x CD-R. My 4x4x NEC CdChanger is the only drive I can't rip with. I use CD Paranoia. It is currently for linux only, and has ripped flawlessly for me, even when using BOTH drives (CDR,DVD) to rip on a P200MMX 128 MB RAM. That's a pretty modest machine. I use it to rip, encode, WHILE using netscape, irc, several terminal sessions, and distributed.net. Sure, my load hovers between 3 and 4, but the machine is usable and doesn't crash.
I currently am encoding with bladeenc, which is much slower than the Xing encoder. It is better at higher bitrates - 160 and above - than the Frauenhauffer(sp) encoder. However, I've been using it at 128 because I find that I still get great sound. I haven't tried the Xing encoder under linux, but perhaps I will today.
You will run across many sites that analyze the quality of mp3s encoded at different bitrates by different encoders. The gist of those sites is this: The higher the bitrate, the better the sound. Nothing beats the Fruenhauffer encoder at 128, but most other encoders aren't noticebly different.
My personal experience is this - if you are encoding so your machine can serve up thousands of mp3s to listen in the background as you work, 128 is fine, and choose the encoder you like.
A great X frontend is gRip - which uses cdparanoia and bladeenc and has cddb capabilities built in. It has debs and rpms if you are looking for ease of installation.
email me with any other questions. miracle@procyon.com
Linux - Because Mommy taught me to Share.