I am aware of the lossy nature of mp3, but since this is how much of audio is available on internet, I want to get the best results I can. I read a recommendation here for certain settings on Audacity, which I applied to saving internet mp3 files and audio from mp4. Based on that, this is what i presently do, and I would like comments on whether this is good or if there is a better way.
I download an mp4 whole, then open it in Audacity at theses settings: 44100 Hz, 32 bit float, best quality sample rates, shaped dither. (If I can’t download the mp4, I record the streaming audio at these settings. Also apply same to downloaded mp3.) Then I edit as needed, change to 16 bit, and export at 320 kbps as WAV. Or should I save as WAV before editing?
The losses of MP3 and MP4 occur during encoding (rather than decoding).
Audacity will always decode MP3 and MP4 the best it can, but some “damage” has already been done by the previous encoding.
The default settings in Audacity are to work in 32 bit float format (uncompressed), however due to limitations in some of the importers, some files will be imported as 16 bit (integer) format. You can see if the imported track in 32 or 16 bit by looking at the track information panel (left end of the track). Before working on an imported track, check whether it says 16 bit or 32 bit float. If it is 16 bit, change it to 32 bit float (click on the track name at the top of the track information panel and select “Set sample format > 32-bit float”. This will ensure that Audacity is working in the highest possible quality.
When exporting, the quality is set by the exporter “options” (Click on the “Options” button in the export dialogue screen).
There is no need to convert to 16 bit before exporting. If that is required the exporter will do that automatically.
320 kbps is the highest standard MP3 setting, but there is still a small amount of sound quality loss due to the encoding.
Before exporting as MP3 (or any other “lossy” compressed format) I would highly recommend exporting as a WAV file so that you have a good quality back-up. If you need to do any further editing or processing of the exported audio you should use either the original MP3/MP4 file and start from scratch, or use the original Audacity project (if you saved the project), or use the exported (backup) WAV file. The important thing is to avoid repeated encoding to a lossy format, because the quality gets a bit worse each time it is encoded.
Thank you, Steve, this is a big help. I want to clarify a couple of things.
I don’t understand about encoding and decoding. When I download an mp3 or mp4, am I encoding or decoding? Which am I doing when I open the file in Audacity? And when I export? Is there any loss during import to Audacity? In the conversion to to WAV? Is there no loss editing a WAV file?
Another poster said we should edit at 32 and export at 16 because “32 bit float is not widely supported by other software”. When you say I don’t need to reset to 16 bit on export, do you mean under edit preferences? Does exporting as 16 bit WAV override the preferences setting?
Processing the way I described and you have qualified, how much damage would you expect starting from a 128 mp3?
No problem, but first let’s clarify about “lossless”.
Strictly speaking “lossless” means that the conversion is “perfect” - no loss of quality at all.
In real life, hardly anything is “perfect”, but in audio engineering there is often “so close to perfect that it makes no audible difference”.
Converting from 32 bit float to 16 bit integer is in this camp. 16 bit WAV format is “very very nearly perfect”, so it is referred to as “lossless”. Mathematically it may not be “absolutely perfect”, but it is so close that it rarely makes any difference.
320 kbps MP3 is “extremely good quality”, but a little less perfect than WAV. Repeatedly re-encoding MP3s, even at 320 kbps, will eventually produce a noticeable deterioration in the sound quality. MP3 is referred to as a “lossy” format because there is always some loss of sound quality.
“Downloading” and “Recording” are different things.
If you are “downloading a file” then you are not doing either encoding or decoding. You are just copying the file from one location (somewhere on the Internet) to another (somewhere on your hard drive).
If you are “recording” an MP3 file (for example, if you are recording Internet radio, then the Internet stream (in MP3 format) is decoded by your radio player and converted by your sound card into analogue sound. You are then converting that analogue sound back into (uncompressed) digital audio and recording it.
“Downloading” a file will produce an “exact” copy of the original file.
“Recording” will produce a slightly reduced quality copy. The quality of the copy depends on how accurately you sound card can do the conversions from digital to analogue and back to digital.
When you “open” (properly called “Importing”) an MP3 file into Audacity, the MP3 is decoded from MP3 into “uncompressed” digital audio. Audacity always works internally with uncompressed audio. The audio in Audacity is an uncompressed (“decoded”) copy of the MP3.
When you Export from Audacity, the audio data is encoded into the required format.
Encoding to WAV format is “lossless”; that is, the encoding is (virtually) perfect.
Encoding to a “lossy” format, such as MP3, MP4, WMA… is “lossy”; that is, there is some loss of sound quality.
Loss of sound quality when encoding to a lossy format is unavoidable. The best we can do is to minimise the amount of deterioration.
No, the audio is decoded and copied (virtually) perfectly from the file into Audacity.
In normal circumstances, working in WAV format is virtually lossless.
For absolute “perfection” 32 bit float format should be used, but for practical reasons it is rarely used outside of Audacity.
Audacity works internally in 32 bit float format (think “EXTREME” high quality).
32 bit float is not widely supported, and file sizes are very large (double the size of a normal 16 bit WAV file).
16 bit WAV is generally considered “close enough” to “perfect” for most purposes.
32 bit float format is useful in audio editors because processing the audio can make small errors much bigger. The extreme quality of 32 bit float format ensures that there is no deterioration in sound quality while you are working.
There are a very few cases where other settings may be slightly preferable, but such cases are quite rare and the possible benefits are small. I would recommend sticking with the default settings here.
I would guess that the additional damage would be hardly noticeable, though of course it is always possible to make a pig’s ear of anything
For real time conversion, the default has medium quality, dither none. Why shouldn’t I use best quality?
I read about dither, but don’t really understand it. Could the dither be desirable when recording imperfect mp3?
I use the files for radio broadcast. I was using mp3 because they will play from a flash drive, WAV will not at our studio. It is much more time and expense to burn to CDs, and there is the problem of labeling on the CDs. Any suggestions? Is there any loss from burning the files to CD?
Those settings only apply if you are playing a track that has a different sample rate to the “Project Rate”.
The “Project Rate” is shown in the bottom left corner of the main Audacity window.
The “Project Rate” determines what sample rate will be used for exported files. Usually this will be 44100.
Even the “medium quality” setting is pretty high quality in absolute terms and is perfectly adequate when listening to Audacity playback.
“Best quality” is even better, but is a bit more demanding on the computer, requiring more processing and make cause stuttering for real time playback, particularly if you are also trying to record additional tracks at the same time.
The short answer is, don’t worry about it, just leave it at the default settings
For the longer answer (may make good bedtime reading if you suffer from insomnia) see: http://wiki.audacityteam.org/wiki/Dither
is that real (air waves) radio, or Internet radio?
I use the files for radio broadcast. I was using mp3 because they will play from a flash drive, WAV will not at our studio.
I think there’s something wrong with your software… All computers should be able to play WAV. Does your software support FLAC (lossless compression)?
Is there any loss from burning the files to CD?
Generally, no. CDs are “lossless”. The underying data is the same uncompressed PCM as a 16-bit, 44.1kHz, stereo WAV file. In theory there is loss if you convert from a high resolution (24/96, etc.). But, CDs are better than human hearing, so it’s generally not an issue.
A couple more comments…
There are special purpose editors such as mp3DirectCut that can do limited editing without decoding (decompressing). mp3DirectCut should also work on AAC/MP4 files. This won’t help if you are recording/capturing streaming audio, but if you are downloading compressed files, you can edit them losslessly.
AAC (MP4) was designed to minimize the “damage” from multiple decode/encode cycles, so editing with a regular audio editor (like Audacity) and then re-saving (exporting) is not as bad as with MP3 (assuming you use the same or higher bitrate when you export).
The general rule is: Don’t do production in a lossy format (whenever possible). If you want to distribute (or stream) a lossy format, compress once as the last step.
The station doesn’t have an on-air computer for playing audio. We have CD players which have flash drive outlets. They will play WAV from CD, but not from flash. Problem was labeling the WAV tracks on CD, and the extra time and expense of CDs, so I was using mp3 on flash. Now I’m exploring WAV on CD anyway due to quality issues.
Thanks for the tip on mp3DirectCut. This may be an option when I have time to learn it.
This is often hardware related and can vary a lot depending on both the drive and the disk being used.
I had one machine that always worked best at the fastest write speed. That is unusual but it does sometimes happen. Probably the best thing is, over a period of time, to try different write speeds and different CD-R brands to see what works best.
Some CD drives are very fussy about the disk being used and only work reliably with certain brands. Other drives can be tolerant and work with almost any disk provided that the disk is not faulty. I’ve not noticed any pattern to which drives / CD-R combinations will work well - it’s usually a case of testing for yourself to see what works for you.
There used to be a free program available from Nero for testing CD quality. I’m not sure if it is still available.
[Update: yes it’s still available. It’s called “Nero Disk Speed” and it’s available here: http://www.nero.com/enu/downloads/ ]
Sooo… Can you confirm what I think I already know. I have a recording over an hour long that was recorded at 32k to fit it onto an ancient mp3 player. I now want to extract a track from it because I can’t find the original single and I want it for my internet stream.
As I see it, if I copy it into Audacity it will make a perfect .au copy of the already lossy recording. I then extract it and turn it back into an mp3, 32k again unfortunately. I will now have an mp3 that will be squished twice to the point where if it only suffers from the swooshles I’ll be really lucky.
Am I right?
This may sound like sacrilege but maybe you should add an option to Audacity that allows editing in the native format especially for this very situation.
Sadly it’s not that easy. Much of Audacity’s ability to do cool things with sound depend on the ability to access the actual (not encoded) audio data. There are other programs that are able to do basic editing (such as splitting and trimming tracks and scaling the volume). Try searching Google for “Split MP3” or similar, but watch out for adware that many of those types of program are often bundled with.
I’ve downloaded seven of them so far. One of them filled the screen with warnings from just about every security program on my computer. Trouble is the first two converted the mp3 into a .wav file, the rest were varying stages of useless (One of them wouldn’t even unpack). The best was mp3DirectCut, but it only seems to be half a program, and doesn’t have a (Working) fade. I suspect I could maybe have used it if the instruction book made sense, but I don’t think English was the programmer’s native language.
I think I’ve wasted enough time on this dead end, it’s back to searching for the 45 again…
It’s not that you can’t do it, it’s just the options are limited. Other than simple editing in an “MP3-Only” editor, you can export at any of the very high quality, uncompressed formats (WAV, AIFF, etc) and what damage there is will get no worse. You can also export as a nose-bleed high MP3. Exporting at 320 Stereo MP3 will minimize the damage increase and still provide an MP3 smaller than the uncompressed formats.
We’re struggling with this problem with another poster. In their case, radio station policy is getting in the way. "We only have provision for show submissions as MP3 or Audio CD. That’s it. So if you have a three-hour show, creating CDs for that could be a career move, and, nobody is clear how high a quality MP3 the system will accept before it crashes.
if you open an MP3 in MP3 Direct Cut, the start and end of the file have a grey square you can drag up and down to adjust the volume envelope. So if you want a 2 seconds fade in for example, you drag the grey square at the start down to the gain required (such as -6 dB gain reduction) then click at 2 seconds then Edit > Create simple fade. There will now be another grey square you can leave at 0 dB gain or whatever gain you want.