|
Post by cynical1 on Mar 20, 2010 17:29:37 GMT -5
Greetings -
Just a quick question to all out there who have encountered sound quality loss when making a conversion from a .wav file to an .mp3.
I've noticed on several mixes that what sounds good in a 44100 16 bit stereo .wav file tends to go South in a New York minute once the trigger is pulled on a 128 bit rate .mp3 conversion.
Is it just me? Anyone found a way around this?
Happy Trails
Cynical One
|
|
|
Post by JohnH on Mar 20, 2010 18:04:46 GMT -5
I use wav2mp3 Wizard: wav2mp3.webternals.com/I reckon that 128rate is the lowest limit for a reasonable quality, and whether it is good enough depends on how sensitive is your hearing and equipment. I find the results are fine, for the purposes for which I need that bit rate, but I try for twice that when I can. John
|
|
|
Post by ashcatlt on Mar 20, 2010 18:55:22 GMT -5
I use a freeware program from NCH called Switch. Not saying it's the best, but different codecs do different things to different sources.
This is a known issue. .mp3 is a lossy "interpretive" compression scheme. Like John said, higher bit rate is always better, but there are some things that can be done to mix to get it to convrt better.
1st, leave some headroom. Make sure the peaks aren't hitting 0dbfs. I've seen songs which peak a half db down end up with overs after conversion.
2- Try to limit or reduce really fast transients. Things like high hats, shakers, etc can kind of confuse the converter.
3 - go over to The TapeOp Message Board and do a search for something like "mixing for mp3" or "...for YouTube" it comes up pretty frequently. It's a little tough for me to search and link from this iPhone.
|
|
|
Post by sumgai on Mar 20, 2010 20:21:14 GMT -5
Allow me to insert my 2 cents worth, if you please..... For years I've been using successive iterations of Any Audio Converter. I probably don't tap more than 2% of what this thing can do, it comes with so much power. Shareware, but the nag screen doesn't seem to get in my way. I follow what ash and John say above, and I concur, but for the greatest part, I don't mix (or re-mix) anything, I just convert whatever I need from what was already recorded somewhere else. From that standpoint, I constantly get things that span a range of dynamics from 86dB to over 100dB. To smooth all that out, I make everything 92dB to start with, using a program called MP3Gain. Freeware I believe, but I could be wrong about that. Audacity also does this, but there's a difference. Audacity works on one file at a time, whereas MP3Gain can compare a whole album (meaning, a whole folder) and modify the files such that the relation between them remains the same, but the overall level can be changed (up or down). Nice feature to have, if you're an album listener. Additionally, you don't have to use the "Album" setting. You can load several folders' worth of tunes at once, and tell it "Single mode". Relativity between tunes is not engaged, and every tune comes out at the average of whatever value you select. This keeps your iPod from smacking you upside the earbones as you travel down the road! ;D And finally, MP3Gain works it's magic with tags, so everything is reversible - you can restore it to the original level, should you so desire. However, MP3Gain does not compress (or expand) anything, whereas Audacity can do that. Another program called MP3Pro (not by the same author) can also do that, but that proggie makes permanent changes - no going back once it's done. (Work on a copy until you're satisfied!) I do use Audacity to trim off blank tails, audience whistling/clapping/etc., but I let it save the changes in the original format (if not already an mp3), then I convert that new file with the aforementioned program. Seems to sound better to my ears, but hanged if I can explain why. HTH! sumgai
|
|
|
Post by cynical1 on Mar 21, 2010 10:19:36 GMT -5
Thanks for all in insight. I've been using Lame Front End to make my conversions. At the higher bitrates it seems to be more then capable. I've grabbed all the tools you guys suggested, and another one called BlueRazorLame, which supports VBR. (variable bit rate) I've been reading up on this and it seems that VBR is something I need to experiment more with. I normally run the bit rate up on any CD's I convert, but my current situation limits me to a 128 bit rate for upload to SoundClick. My gut tells me Ash hit it on the head with the levels on my mix...along with my rabid use of fast transients...in .wav format when I convert it to .mp3. As a rule I'm happy with the master when the peak is around -.2 to -.5 db and the RMS is somewhere around -10 db. Granted, this is not a holy grail by any means, but it's where I always ran the analog stuff back in the Dark Ages...and old habits die hard... I'm still adjusting to the fact I have no physical knobs or sliders to work with in this digital world... I will play with this stuff and report back...or come back with more WTF moments... Happy Trails Cynical One
|
|
|
Post by dunkelfalke on Mar 21, 2010 11:15:56 GMT -5
as long as you are using mp3, the lame codec is the best you can get anyway.
|
|
|
Post by ashcatlt on Mar 21, 2010 11:51:01 GMT -5
Boy, cyn, with levels like that you could be a professional! I'm not quite sure what type of music you're talking about, but you've got to be coming close to total slamdamage here. The peak levels would be fine for .wav or CD release. I might not go quite as high as -0.2 even in that case, because some consumer gear can end up clipping at this level. -10 for RMS seems a bit high though. There are plenty of commercial releases out there with this kind of dynamic range, but they're not necessarily the best sounding discs out there. I personally try to stay down below -12, and usually closer to -15. This is, of course, excepting some of my noise experiments. So, are you the guy who started the Loudness Wars? I had the thought that you might try some low-pass filtering. On the mix as a whole and/or on some of the "faster" tracks. This might sound counter-intuitive, but ever so slightly knocking down the very highest frequencies on something like a hi-hat won't generally come across as a loss of treble, and just might make it play a little better with your encoder. I always "bookend" my master fader with an LPF set between 15K and 18K and an HPF somewhere down between 20 and 40 Hz. The HPF gets rid of any unwanted subharmonics which can eat up headroom without actually being heard. Digital can reproduce frequencies all the way down to 0Hz. Your listeners aren't likely to hear anything much below 40 even on relatively high-end consumer gear. The LPF is mostly to "help" the anti-aliasing filter which is a very steep filter up around 22kHz (I always record at 44.1). Being a very steep filter, the anti-alias can cause some strange things to happen in the area close to its cutoff. The less it has to work on, the less trouble it can cause. I've been doing it ever since I started this digital recording thing, and was somewhat vindicated when I read several actual working (like for real money) engineers say they do it too. This area of the extreme high frequencies is outside the range of what we generally call treble, and much more like just the "air" in the mix. It also happens to be the area where the "digital" lives. Anyway, for your situation you might even have to set the LPF a little lower, maybe even 12K, but it might not have to be quite so steep: Q = 1 or 1.5. It probably will sound a little darker, but it could encode better. Now, I've never really had this trouble. Maybe I'm just lucky, but I'm usually pretty satisfied with the .mp3 versions of my mixes. At a certain point, you're really just going to have to live with some loss at this bitrate. If you're sure that you're happy with the mix in .wav format, there's no good reason to make drastic changes just for .mp3.
|
|
|
Post by cynical1 on Mar 22, 2010 7:29:28 GMT -5
I agree with the lpf around 15khz. Radio transmitters cut it at 15khz, so I figure why shouldn't I... I also generally run a hpf with at -12db or better for around 40-60 hz. Again, why waste the bandwidth if it isn't going to be heard...
I've never held that RMS is an end all be all, it's just a good yardstick for me. I'll try the kinder and more gentle -12 to -15 db.
And no, I'm not the guy who started the loudness wars...although, if they ain't bleeding from the eyes out to the third row it ain't loud enough... Actually, I generally run a compressed and non-compressed drum track together to allow the non-compressed track to preserve some dynamics...rather then squash it down to a blur...
I've got one song now I'm working on with three guitar tracks, bass, drums, horns and vocals. All sounds fine in Sonar at the master, but I just don't like what I hear when I squash it down to an .mp3. I'm gonna try some things in the mix, then try the VBR conversion and see if I like it any better.
I admit, there are things about recording, sequencing and mixing digitally that I just love. There are other things that I just can't get a handle on...but, this too shall pass...
Happy Trails
Cynical One
|
|
|
Post by JFrankParnell on May 28, 2010 18:41:41 GMT -5
Here's some links to the best mp3 settings and programs. wiki.hydrogenaudio.org/index.php?title=EAC_and_Lamewww.hydrogenaudio.org/forums/index.php?showtopic=28124That hydrogenaudio is your best source of mp3 info. Basically, your 'best' settings are going to be VBR, averaging 192kbs or so, in joint stereo. Can you hear the difference between your wav and your mp3? You shouldnt be able to, and if you can, it's because either, A. you havent converted at the 'best' possible settings or B. actually, you cant ;D and you can test your hearing/encoding with this: ff123.net/abchr/abchr.html Its a blind listening test. I bet my buddy $100 he couldnt tell the difference. After a couple AB trials, he wouldnt take the bet ;D
|
|
|
Post by cynical1 on May 28, 2010 21:14:56 GMT -5
jfrank -
you're dead on with the quality you can get in an .mp3 with no bandwidth restrictions. The problem I have now is that Soundclick mandate 128 as a brickwall for uploads on free accounts.
I've been playing with LFE and the last attempt on Playing in Traffic is about the best. I did find that if I removed the 44100 mandate and let it drop to 33000 it preserved as much character as possible and worked out pretty well for streaming Internet audio.
I personally prefer .ogg or .flac files...but Soundclick doesn't allow you to upload that for free...
Happy Trails
Cynical One
|
|
|
Post by ashcatlt on May 28, 2010 21:46:39 GMT -5
If any reasonable software is converting your files down to 33K, it must be applying a pretty severe low-pass (anti-aliasing) filter at or below 16.5K. Sounds alot like something we said earlier in this thread... Of course, it opens up the new question of the evils of sample-rate conversion, which can be really nasty. I've honestly never tried it where I wasn't specifically looking to mess something up, but I've seen pictures.
|
|
|
Post by JFrankParnell on May 29, 2010 9:19:03 GMT -5
Cynical1, I listened to Legal Fiction, cool stuff! I didnt really critically listen though.
Is LFE an acronym for Legal Fiction or a program? Oh, Lame Free Encoder? Well, anyway, are you really letting sample rate drop or bit rate? I agree, Ashcat, sample rate conversion from 44.1k>32k is not recomended.
If you need to encode at 128kbs (damn you, Soundclick) I would suggest "-b 128" as the start of the command line sent to LAME. With the latest recomended LAME, etc. I suggest EAC or Lame Front End or a program that lets you use the latest Lame and lets you specify the command line, as both of these things together should result in the best mp3.
|
|